{"schema":"5lc-index-ledger/1","key":"models-specialist-models","layer":"models","sublayer":"specialist-models","asOf":"2026-10-22","about":"5LC Importance Index ledger for one sublayer: the measure card, the current ten with any editorial overrides and their reasons, the computed result for every scored company (Importance = share of AI output × years to replace, Reach, replacement years, Rising, the Monte Carlo rank simulation and its stability class, validation, the revenue baseline and rater agreement), each scored company's sourced inputs and irreplaceability ratings, and the sublayer's validation cases with their sources. Monte Carlo frequencies describe the stated input ranges, not true probabilities. Omitted: the free-text notes of the internal check passes and of rating reconciliations, and the location of the internal card review; references to local evidence files and to internal method decisions are replaced by plain descriptions.","roster":{"members":[{"entityId":"bytedance-seed","slot":"importance","since":"2026-10-22"},{"entityId":"elevenlabs","slot":"importance","since":"2026-10-22"},{"entityId":"suno","slot":"importance","since":"2026-10-22"},{"entityId":"google-deepmind","slot":"importance","since":"2026-10-22"},{"entityId":"kling-ai","slot":"importance","since":"2026-10-22"},{"entityId":"midjourney","slot":"importance","since":"2026-10-22"},{"entityId":"adobe-firefly","slot":"importance","since":"2026-10-22"},{"entityId":"openai","slot":"importance","since":"2026-10-22"},{"entityId":"skild-ai","slot":"rising","since":"2026-10-22"},{"entityId":"heygen","slot":"rising","since":"2026-10-22"}],"pending":{},"lastRun":"2026-10-22","resultsDate":"2026-10-22"},"card":{"schema":"5lc-index-card/1","key":"models-specialist-models","layer":"models","sublayer":"specialist-models","version":1,"scope":"Models built for one kind of output (speech and voice, music, images and video, avatar video, robot control, and molecule and materials design), counted for their developer, including such models from large companies such as Google, OpenAI, ByteDance, Microsoft, Amazon and Adobe. Stand-alone specialist products and specialist-model APIs and licences count (for example Kling, Midjourney, Suno, Firefly plans and the Veo API); specialist features inside general assistants or creative suites (ChatGPT, the Gemini app, Doubao, Creative Cloud) count in Applications. General-purpose and coding models count in Frontier labs or Open-weight, hosting of others' models in Inference serving, apps built on other companies' models in Applications, and text-translation models are not scored in this version.","outputMeasure":"Revenue that specialist models' developers earn from their output: voices, music, images, video, robot skills and molecule designs.","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","hardToReplace":"Most specialist outputs have several capable models, and an app can switch within weeks. Harder to replace are voice and avatar models built into business workflows, and robot and science models, where few developers have the data and industry partners at scale.","functions":[{"id":"speech-voice","name":"Speech and voice","why":"Text-to-speech, transcription and voice-agent models.","weight":{"low":0.129,"central":0.239,"high":0.433},"weightSource":{"url":"https://techcrunch.com/2026/09/24/twenty-minutes-with-the-ceo-of-elevenlabs-now-reportedly-valued-at-22-billion/","title":"ElevenLabs' CEO on margins, IPO timing, and telling customers they're talking to a bot","publisher":"TechCrunch","publishedDate":"2026-09-24","accessedDate":"2026-10-22","tier":"T2","locator":"The company says (CEO Mati Staniszewski): \"We are now $600 million in ARR\"; 55% or more from enterprise customers"},"shareData":"Company ARR statements and press reports; labelled estimates for large companies' speech services"},{"id":"music","name":"Music","why":"Song-generation models, sold mostly as creator subscriptions.","weight":{"low":0.034,"central":0.056,"high":0.119},"weightSource":{"url":"https://techcrunch.com/2026/02/27/ai-music-generator-suno-hits-2-million-paid-subscribers-and-300m-in-annual-recurring-revenue/","title":"AI music generator Suno hits 2M paid subscribers and $300M in annual recurring revenue","publisher":"TechCrunch","publishedDate":"2026-02-27","accessedDate":"2026-10-22","tier":"T2","locator":"The company says (CEO Mikey Shulman on LinkedIn): 2 million paid subscribers and $300 million ARR; $200 million annual revenue three months earlier"},"shareData":"Company ARR and subscriber statements"},{"id":"image-video","name":"Images and video","why":"Image and video generation for creators, advertisers and studios; most of the revenue.","weight":{"low":0.367,"central":0.597,"high":0.749},"weightSource":{"url":"https://www1.hkexnews.hk/listedco/listconews/sehk/2026/0819/2026081900314.pdf","title":"Kuaishou Technology results announcement for the three and six months ended June 30, 2026","publisher":"Kuaishou Technology (HKEXnews)","publishedDate":"2026-08-19","accessedDate":"2026-10-22","tier":"T1","locator":"p. 5, AI business: Kling AI generated revenue of over RMB850 million in the second quarter of 2026, up more than 200% year over year"},"shareData":"Kling and Firefly disclosures, company statements and press reports; labelled estimates for large companies' units"},{"id":"avatar-video","name":"Avatar video","why":"Presenter and translated videos made for business training, sales and marketing.","weight":{"low":0.043,"central":0.075,"high":0.132},"weightSource":{"url":"https://www.heygen.com/blog/heygen-surpasses-200m-arr","title":"HeyGen Doubles to $200M ARR in Eight Months on the Rise of Identity-First AI Video","publisher":"HeyGen","publishedDate":"2026-06-25","accessedDate":"2026-10-22","tier":"T2","locator":"HeyGen has surpassed $200 million in annual recurring revenue, doubling in eight months"},"shareData":"Company ARR statements and press reports"},{"id":"robotics-science","name":"Robot and science models","why":"Models that control robots or design molecules; small revenue so far.","weight":{"low":0.015,"central":0.033,"high":0.083},"weightSource":{"url":"https://www.skild.ai/blogs/skild-crosses-100m-arr","title":"The Hidden Pillar of Robotics","publisher":"Skild AI","publishedDate":"2026-09-10","accessedDate":"2026-10-22","tier":"T2","locator":"Above $100 million ARR; $50 million recognised in about 10 months of deployments; more than 60 paying customers; about 90% manipulation, 10% mobility"},"shareData":"Company revenue statements and disclosed partnership payments"}],"risingFloorPct":1,"validationCaseIds":[],"knownGaps":["Large companies do not disclose revenue for their specialist models (Google Veo, Imagen and speech services; OpenAI's image and audio APIs; Microsoft and Amazon speech services; ByteDance Seedance and Seedream), so the weights include 5LayerCake estimates for them; Adobe reports Firefly ARR.","Midjourney and Black Forest Labs publish no current revenue; their sizes are estimates built from older company and press figures.","Midjourney's low end (0.30) rests on TIME's reported $300m for 2024; the stated estimate method alone would give 0.24.","Science models earn mostly through partnership payments that are not disclosed.","Companies define ARR differently, and the latest figures range from February to September 2026.","Code models have no part here: closed code models count in Frontier labs' coding part, open ones in Open-weight, and coding tools in Applications · Developer tools.","Text-translation models (for example DeepL) have no current public revenue data and are not scored in this version.","Creation apps are also measured in Applications · Consumer, by monthly users (an allowed separate placement); HeyGen, Synthesia and Runway are counted here and not in Vertical AI."],"approval":{"approver":"Claude (delegated by Dachi)","date":"2026-10-22","sha256":"6c082ebb84e7477ce73639566cfb841577353c1b6210bd8454e9d1adf045aa63","delegatedOn":"2026-10-22"}},"result":{"schema":"5lc-index-result/1","key":"models-specialist-models","asOf":"2026-10-22","methodVersion":"1.1","methodSha256":"3cabed6b38514c263d8b86a494e8cc19cdbd0b2a1013543dc024192c49f7a6a6","cardSha256":"6c082ebb84e7477ce73639566cfb841577353c1b6210bd8454e9d1adf045aa63","validationSha256":"409931bb458f38b02342c9bf43e61eab299e47d8679912af2c96074ec556f17e","inputsSha256":"34f6965e4338ce6f145795f59023fca9c66075b978f7c96e92d8c112a39d24bc","unscreened":[],"companies":[{"entityId":"bytedance-seed","name":"ByteDance Seed (Seedance, Seedream)","reachPct":14.7228255,"importance":4.59479869866,"importanceRank":1,"functions":[{"id":"speech-voice","weight":0.239,"sharePct":2.1291,"shareLowPct":0.5323,"shareHighPct":7.9842,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false},{"id":"music","weight":0.056,"sharePct":2.0817,"shareLowPct":0.5204,"shareHighPct":7.8064,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false},{"id":"image-video","weight":0.597,"sharePct":23.5632,"shareLowPct":9.4253,"shareHighPct":42.4137,"stale":false,"offBasis":false,"level":1.33333333333,"years":0.3149802624729905,"disagreement":false},{"id":"avatar-video","weight":0.075,"sharePct":0.4022,"shareLowPct":0,"shareHighPct":3.0163,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"robotics-science","weight":0.033,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false}],"rising":null,"risingEligible":true,"risingComponents":{},"mc":{"medianRank":1,"rankP10":1,"rankP90":3,"pTop8":1,"pSelected":1,"stability":"firm"},"baselineRevenueRank":null},{"entityId":"elevenlabs","name":"ElevenLabs","reachPct":7.8515222,"importance":3.45473436716,"importanceRank":2,"functions":[{"id":"speech-voice","weight":0.239,"sharePct":31.937,"shareLowPct":28.7433,"shareHighPct":35.1307,"stale":false,"offBasis":false,"level":1.83333333333,"years":0.4454493590691404,"disagreement":false},{"id":"music","weight":0.056,"sharePct":3.9032,"shareLowPct":1.3011,"shareHighPct":10.4085,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false},{"id":"image-video","weight":0.597,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"avatar-video","weight":0.075,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"robotics-science","weight":0.033,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false}],"rising":75,"risingEligible":true,"risingComponents":{"growth":66.6666666667,"shareChange":83.3333333333},"mc":{"medianRank":2,"rankP10":1,"rankP90":4,"pTop8":0.9975,"pSelected":1,"stability":"firm"},"baselineRevenueRank":1},{"entityId":"suno","name":"Suno","reachPct":4.371584,"importance":2.4534685655,"importanceRank":3,"functions":[{"id":"speech-voice","weight":0.239,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"music","weight":0.056,"sharePct":78.064,"shareLowPct":70.2576,"shareHighPct":85.8704,"stale":false,"offBasis":false,"level":2.16666666667,"years":0.5612310241559831,"disagreement":false},{"id":"image-video","weight":0.597,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"avatar-video","weight":0.075,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"robotics-science","weight":0.033,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false}],"rising":null,"risingEligible":true,"risingComponents":{},"mc":{"medianRank":3,"rankP10":2,"rankP90":7,"pTop8":0.9698,"pSelected":0.9855,"stability":"firm"},"baselineRevenueRank":4},{"entityId":"google-deepmind","name":"Google DeepMind","reachPct":8.6762117,"importance":2.169052925,"importanceRank":4,"functions":[{"id":"speech-voice","weight":0.239,"sharePct":9.0474,"shareLowPct":3.1932,"shareHighPct":21.2879,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false},{"id":"music","weight":0.056,"sharePct":2.0833,"shareLowPct":0.5208,"shareHighPct":7.8125,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false},{"id":"image-video","weight":0.597,"sharePct":10.6431,"shareLowPct":3.1929,"shareHighPct":23.947,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false},{"id":"avatar-video","weight":0.075,"sharePct":0.2011,"shareLowPct":0,"shareHighPct":2.0109,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"robotics-science","weight":0.033,"sharePct":0.8547,"shareLowPct":0,"shareHighPct":4.2735,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true}],"rising":null,"risingEligible":true,"risingComponents":{},"mc":{"medianRank":4,"rankP10":2,"rankP90":6,"pTop8":0.9739,"pSelected":0.9908,"stability":"firm"},"baselineRevenueRank":null},{"entityId":"kling-ai","name":"Kling AI","reachPct":7.9381896,"importance":1.9845474,"importanceRank":5,"functions":[{"id":"speech-voice","weight":0.239,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"music","weight":0.056,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"image-video","weight":0.597,"sharePct":13.2968,"shareLowPct":12.63196,"shareHighPct":13.96164,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false},{"id":"avatar-video","weight":0.075,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"robotics-science","weight":0.033,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false}],"rising":91.6666666667,"risingEligible":true,"risingComponents":{"growth":83.3333333333,"shareChange":100},"mc":{"medianRank":6,"rankP10":4,"rankP90":8,"pTop8":0.9611,"pSelected":1,"stability":"firm"},"baselineRevenueRank":2},{"entityId":"midjourney","name":"Midjourney","reachPct":6.354468,"importance":1.7831622918,"importanceRank":6,"functions":[{"id":"speech-voice","weight":0.239,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"music","weight":0.056,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"image-video","weight":0.597,"sharePct":10.644,"shareLowPct":7.983,"shareHighPct":14.6354,"stale":false,"offBasis":false,"level":1.16666666667,"years":0.2806155120779916,"disagreement":false},{"id":"avatar-video","weight":0.075,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"robotics-science","weight":0.033,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false}],"rising":null,"risingEligible":true,"risingComponents":{},"mc":{"medianRank":7,"rankP10":4,"rankP90":9,"pTop8":0.8378,"pSelected":0.9241,"stability":"firm"},"baselineRevenueRank":3},{"entityId":"adobe-firefly","name":"Adobe Firefly","reachPct":4.371187,"importance":1.72267657546,"importanceRank":7,"functions":[{"id":"speech-voice","weight":0.239,"sharePct":0.1597,"shareLowPct":0,"shareHighPct":1.0644,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"music","weight":0.056,"sharePct":0.7813,"shareLowPct":0,"shareHighPct":5.2083,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"image-video","weight":0.597,"sharePct":7.1847,"shareLowPct":4.3906,"shareHighPct":9.5796,"stale":false,"offBasis":false,"level":1.66666666667,"years":0.3968502629929668,"disagreement":false},{"id":"avatar-video","weight":0.075,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"robotics-science","weight":0.033,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false}],"rising":null,"risingEligible":true,"risingComponents":{},"mc":{"medianRank":8,"rankP10":5,"rankP90":11,"pTop8":0.6135,"pSelected":0.7565,"stability":"likely"},"baselineRevenueRank":5},{"entityId":"openai","name":"OpenAI (image, video and audio models)","reachPct":5.0256304,"importance":1.51445619814,"importanceRank":8,"functions":[{"id":"speech-voice","weight":0.239,"sharePct":4.2583,"shareLowPct":1.0646,"shareHighPct":13.3071,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false},{"id":"music","weight":0.056,"sharePct":0.2602,"shareLowPct":0,"shareHighPct":2.6021,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"image-video","weight":0.597,"sharePct":6.6519,"shareLowPct":1.5965,"shareHighPct":15.9647,"stale":false,"offBasis":false,"level":1.33333333333,"years":0.3149802624729905,"disagreement":false},{"id":"avatar-video","weight":0.075,"sharePct":0.2011,"shareLowPct":0,"shareHighPct":2.0109,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"robotics-science","weight":0.033,"sharePct":0.2139,"shareLowPct":0,"shareHighPct":2.1395,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true}],"rising":null,"risingEligible":true,"risingComponents":{},"mc":{"medianRank":7,"rankP10":4,"rankP90":11,"pTop8":0.744,"pSelected":0.8419,"stability":"likely"},"baselineRevenueRank":null},{"entityId":"heygen","name":"HeyGen","reachPct":3.018105,"importance":1.19773576299,"importanceRank":9,"functions":[{"id":"speech-voice","weight":0.239,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"music","weight":0.056,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"image-video","weight":0.597,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"avatar-video","weight":0.075,"sharePct":40.2414,"shareLowPct":36.21726,"shareHighPct":44.26554,"stale":false,"offBasis":false,"level":1.66666666667,"years":0.3968502629929668,"disagreement":false},{"id":"robotics-science","weight":0.033,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false}],"rising":50,"risingEligible":true,"risingComponents":{"growth":50,"shareChange":50},"mc":{"medianRank":10,"rankP10":7,"rankP90":13,"pTop8":0.2119,"pSelected":0.9956,"stability":"firm"},"baselineRevenueRank":7},{"entityId":"skild-ai","name":"Skild AI","reachPct":1.4120667,"importance":1.12075808252,"importanceRank":10,"functions":[{"id":"speech-voice","weight":0.239,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"music","weight":0.056,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"image-video","weight":0.597,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"avatar-video","weight":0.075,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"robotics-science","weight":0.033,"sharePct":42.7899,"shareLowPct":38.51091,"shareHighPct":47.06889,"stale":false,"offBasis":false,"level":2.66666666667,"years":0.7937005259859335,"disagreement":false}],"rising":83.3333333333,"risingEligible":true,"risingComponents":{"growth":100,"shareChange":66.6666666667},"mc":{"medianRank":9,"rankP10":5,"rankP90":14,"pTop8":0.3768,"pSelected":1,"stability":"firm"},"baselineRevenueRank":12},{"entityId":"synthesia","name":"Synthesia","reachPct":2.2622175,"importance":1.00770333545,"importanceRank":11,"functions":[{"id":"speech-voice","weight":0.239,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"music","weight":0.056,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"image-video","weight":0.597,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"avatar-video","weight":0.075,"sharePct":30.1629,"shareLowPct":27.14661,"shareHighPct":33.17919,"stale":false,"offBasis":false,"level":1.83333333333,"years":0.4454493590691404,"disagreement":false},{"id":"robotics-science","weight":0.033,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false}],"rising":33.3333333333,"risingEligible":true,"risingComponents":{"growth":33.3333333333,"shareChange":33.3333333333},"mc":{"medianRank":12,"rankP10":9,"rankP90":15,"pTop8":0.0848,"pSelected":0.1637,"stability":"borderline"},"baselineRevenueRank":9},{"entityId":"runway","name":"Runway","reachPct":3.1769952,"importance":0.7942488,"importanceRank":12,"functions":[{"id":"speech-voice","weight":0.239,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"music","weight":0.056,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"image-video","weight":0.597,"sharePct":5.3216,"shareLowPct":4.78944,"shareHighPct":5.85376,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false},{"id":"avatar-video","weight":0.075,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"robotics-science","weight":0.033,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false}],"rising":null,"risingEligible":true,"risingComponents":{},"mc":{"medianRank":14,"rankP10":12,"rankP90":16,"pTop8":0.0003,"pSelected":0.0016,"stability":"borderline"},"baselineRevenueRank":8},{"entityId":"microsoft-ai","name":"Microsoft AI (MAI) and Azure AI Speech","reachPct":2.6808764,"importance":0.6702191,"importanceRank":13,"functions":[{"id":"speech-voice","weight":0.239,"sharePct":9.0488,"shareLowPct":3.1937,"shareHighPct":21.2913,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false},{"id":"music","weight":0.056,"sharePct":0.1301,"shareLowPct":0,"shareHighPct":1.3011,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"image-video","weight":0.597,"sharePct":0.7982,"shareLowPct":0.133,"shareHighPct":2.6608,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"avatar-video","weight":0.075,"sharePct":0.4022,"shareLowPct":0,"shareHighPct":2.413,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"robotics-science","weight":0.033,"sharePct":0.1284,"shareLowPct":0,"shareHighPct":1.2837,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true}],"rising":null,"risingEligible":true,"risingComponents":{},"mc":{"medianRank":12,"rankP10":9,"rankP90":16,"pTop8":0.0884,"pSelected":0.1288,"stability":"borderline"},"baselineRevenueRank":null},{"entityId":"soundhound","name":"SoundHound AI","reachPct":1.496618,"importance":0.666667528871,"importanceRank":14,"functions":[{"id":"speech-voice","weight":0.239,"sharePct":6.262,"shareLowPct":3.1587,"shareHighPct":13.1766,"stale":false,"offBasis":false,"level":1.83333333333,"years":0.4454493590691404,"disagreement":false},{"id":"music","weight":0.056,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"image-video","weight":0.597,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"avatar-video","weight":0.075,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"robotics-science","weight":0.033,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false}],"rising":22.2222222222,"risingEligible":true,"risingComponents":{"growth":16.6666666667,"shareChange":0,"committed":50},"mc":{"medianRank":14,"rankP10":9,"rankP90":18,"pTop8":0.0777,"pSelected":0.1039,"stability":"borderline"},"baselineRevenueRank":6},{"entityId":"iflytek","name":"iFlytek (speech)","reachPct":1.2735207,"importance":0.587431025072,"importanceRank":15,"functions":[{"id":"speech-voice","weight":0.239,"sharePct":4.4426,"shareLowPct":1.11065,"shareHighPct":10.36595,"stale":false,"offBasis":true,"level":2,"years":0.5,"disagreement":false},{"id":"music","weight":0.056,"sharePct":0.2602,"shareLowPct":0,"shareHighPct":1.5613,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"image-video","weight":0.597,"sharePct":0.133,"shareLowPct":0.0266,"shareHighPct":0.7982,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"avatar-video","weight":0.075,"sharePct":1.0054,"shareLowPct":0.2011,"shareHighPct":5.0271,"stale":false,"offBasis":false,"level":1.16666666667,"years":0.2806155120779916,"disagreement":false},{"id":"robotics-science","weight":0.033,"sharePct":1.2837,"shareLowPct":0,"shareHighPct":8.558,"stale":false,"offBasis":false,"level":1.16666666667,"years":0.2806155120779916,"disagreement":false}],"rising":8.33333333335,"risingEligible":true,"risingComponents":{"growth":0,"shareChange":16.6666666667},"mc":{"medianRank":14,"rankP10":10,"rankP90":18,"pTop8":0.0291,"pSelected":0.0461,"stability":"borderline"},"baselineRevenueRank":null},{"entityId":"amazon-agi","name":"Amazon Nova media models, Polly and Transcribe","reachPct":2.1972127,"importance":0.549303175,"importanceRank":16,"functions":[{"id":"speech-voice","weight":0.239,"sharePct":7.452,"shareLowPct":2.6614,"shareHighPct":18.6299,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false},{"id":"music","weight":0.056,"sharePct":0.1301,"shareLowPct":0,"shareHighPct":1.3011,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"image-video","weight":0.597,"sharePct":0.6652,"shareLowPct":0.133,"shareHighPct":2.1286,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"avatar-video","weight":0.075,"sharePct":0.1005,"shareLowPct":0,"shareHighPct":1.0054,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"robotics-science","weight":0.033,"sharePct":0.1284,"shareLowPct":0,"shareHighPct":1.2837,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true}],"rising":null,"risingEligible":true,"risingComponents":{},"mc":{"medianRank":14,"rankP10":10,"rankP90":17,"pTop8":0.0298,"pSelected":0.0492,"stability":"borderline"},"baselineRevenueRank":null},{"entityId":"black-forest-labs","name":"Black Forest Labs","reachPct":1.5296931,"importance":0.382423275,"importanceRank":17,"functions":[{"id":"speech-voice","weight":0.239,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"music","weight":0.056,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"image-video","weight":0.597,"sharePct":2.5623,"shareLowPct":1.02492,"shareHighPct":4.09968,"stale":true,"offBasis":false,"level":1,"years":0.25,"disagreement":false},{"id":"avatar-video","weight":0.075,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"robotics-science","weight":0.033,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false}],"rising":null,"risingEligible":true,"risingComponents":{},"mc":{"medianRank":21,"rankP10":17,"rankP90":27,"pTop8":0,"pSelected":0,"stability":"borderline"},"baselineRevenueRank":13},{"entityId":"field-ai","name":"Field AI","reachPct":0.6282705,"importance":0.352604896162,"importanceRank":18,"functions":[{"id":"speech-voice","weight":0.239,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"music","weight":0.056,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"image-video","weight":0.597,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"avatar-video","weight":0.075,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"robotics-science","weight":0.033,"sharePct":19.0385,"shareLowPct":6.3462,"shareHighPct":42.6923,"stale":false,"offBasis":false,"level":2.16666666667,"years":0.5612310241559831,"disagreement":false}],"rising":null,"risingEligible":false,"risingComponents":{},"mc":{"medianRank":18,"rankP10":13,"rankP90":27,"pTop8":0.0034,"pSelected":0.0117,"stability":"borderline"},"baselineRevenueRank":null},{"entityId":"deepgram","name":"Deepgram","reachPct":1.271958,"importance":0.3179895,"importanceRank":19,"functions":[{"id":"speech-voice","weight":0.239,"sharePct":5.322,"shareLowPct":3.1932,"shareHighPct":7.4508,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false},{"id":"music","weight":0.056,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"image-video","weight":0.597,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"avatar-video","weight":0.075,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"robotics-science","weight":0.033,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false}],"rising":null,"risingEligible":true,"risingComponents":{},"mc":{"medianRank":21,"rankP10":18,"rankP90":27,"pTop8":0,"pSelected":0,"stability":"borderline"},"baselineRevenueRank":10},{"entityId":"topaz-labs","name":"Topaz Labs","reachPct":0.7147881,"importance":0.283663845469,"importanceRank":20,"functions":[{"id":"speech-voice","weight":0.239,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"music","weight":0.056,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"image-video","weight":0.597,"sharePct":1.1973,"shareLowPct":0.2661,"shareHighPct":3.1929,"stale":false,"offBasis":false,"level":1.66666666667,"years":0.3968502629929668,"disagreement":false},{"id":"avatar-video","weight":0.075,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"robotics-science","weight":0.033,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false}],"rising":null,"risingEligible":false,"risingComponents":{},"mc":{"medianRank":23,"rankP10":17,"rankP90":33,"pTop8":0,"pSelected":0,"stability":"borderline"},"baselineRevenueRank":null},{"entityId":"assemblyai","name":"AssemblyAI","reachPct":0.9668506,"importance":0.24171265,"importanceRank":21,"functions":[{"id":"speech-voice","weight":0.239,"sharePct":4.0454,"shareLowPct":2.42724,"shareHighPct":5.66356,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false},{"id":"music","weight":0.056,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"image-video","weight":0.597,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"avatar-video","weight":0.075,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"robotics-science","weight":0.033,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false}],"rising":null,"risingEligible":false,"risingComponents":{},"mc":{"medianRank":25,"rankP10":20,"rankP90":31,"pTop8":0,"pSelected":0,"stability":"borderline"},"baselineRevenueRank":15},{"entityId":"xai","name":"xAI (Grok Imagine)","reachPct":0.9658253,"importance":0.241456325,"importanceRank":22,"functions":[{"id":"speech-voice","weight":0.239,"sharePct":0.0532,"shareLowPct":0,"shareHighPct":0.5322,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"music","weight":0.056,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"image-video","weight":0.597,"sharePct":1.5965,"shareLowPct":0.2661,"shareHighPct":6.6519,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false},{"id":"avatar-video","weight":0.075,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"robotics-science","weight":0.033,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false}],"rising":null,"risingEligible":false,"risingComponents":{},"mc":{"medianRank":19,"rankP10":15,"rankP90":31,"pTop8":0.0001,"pSelected":0.0001,"stability":"borderline"},"baselineRevenueRank":null},{"entityId":"meshy","name":"Meshy","reachPct":0.7943085,"importance":0.222895286475,"importanceRank":23,"functions":[{"id":"speech-voice","weight":0.239,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"music","weight":0.056,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"image-video","weight":0.597,"sharePct":1.3305,"shareLowPct":0,"shareHighPct":2.661,"stale":false,"offBasis":false,"level":1.16666666667,"years":0.2806155120779916,"disagreement":false},{"id":"avatar-video","weight":0.075,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"robotics-science","weight":0.033,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false}],"rising":null,"risingEligible":false,"risingComponents":{},"mc":{"medianRank":29,"rankP10":21,"rankP90":37,"pTop8":0,"pSelected":0,"stability":"borderline"},"baselineRevenueRank":11},{"entityId":"photoroom","name":"Photoroom","reachPct":0.88953,"importance":0.2223825,"importanceRank":24,"functions":[{"id":"speech-voice","weight":0.239,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"music","weight":0.056,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"image-video","weight":0.597,"sharePct":1.49,"shareLowPct":0.5002,"shareHighPct":2.9588,"stale":true,"offBasis":false,"level":1,"years":0.25,"disagreement":false},{"id":"avatar-video","weight":0.075,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"robotics-science","weight":0.033,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false}],"rising":null,"risingEligible":false,"risingComponents":{},"mc":{"medianRank":27,"rankP10":21,"rankP90":34,"pTop8":0,"pSelected":0,"stability":"borderline"},"baselineRevenueRank":14},{"entityId":"pixverse","name":"AIsphere (PixVerse)","reachPct":0.7943085,"importance":0.198577125,"importanceRank":25,"functions":[{"id":"speech-voice","weight":0.239,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"music","weight":0.056,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"image-video","weight":0.597,"sharePct":1.3305,"shareLowPct":1.0644,"shareHighPct":1.9957,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false},{"id":"avatar-video","weight":0.075,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"robotics-science","weight":0.033,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false}],"rising":null,"risingEligible":false,"risingComponents":{},"mc":{"medianRank":29,"rankP10":25,"rankP90":33,"pTop8":0,"pSelected":0,"stability":"borderline"},"baselineRevenueRank":17},{"entityId":"alibaba-qwen","name":"Alibaba Qwen (Wan, Qwen-Image, speech)","reachPct":0.7817094,"importance":0.19542735,"importanceRank":26,"functions":[{"id":"speech-voice","weight":0.239,"sharePct":0.5322,"shareLowPct":0.1064,"shareHighPct":2.1288,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"music","weight":0.056,"sharePct":0.1302,"shareLowPct":0,"shareHighPct":1.3021,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"image-video","weight":0.597,"sharePct":1.0644,"shareLowPct":0.2129,"shareHighPct":3.9915,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false},{"id":"avatar-video","weight":0.075,"sharePct":0.1006,"shareLowPct":0,"shareHighPct":1.006,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"robotics-science","weight":0.033,"sharePct":0.1282,"shareLowPct":0,"shareHighPct":1.2821,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true}],"rising":null,"risingEligible":false,"risingComponents":{},"mc":{"medianRank":22,"rankP10":17,"rankP90":30,"pTop8":0,"pSelected":0,"stability":"borderline"},"baselineRevenueRank":null},{"entityId":"minimax","name":"MiniMax (Hailuo, Speech, Music)","reachPct":0.7374081,"importance":0.184352025,"importanceRank":27,"functions":[{"id":"speech-voice","weight":0.239,"sharePct":0.7984,"shareLowPct":0.0133,"shareHighPct":2.5949,"stale":false,"offBasis":true,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"music","weight":0.056,"sharePct":1.0409,"shareLowPct":0.2602,"shareHighPct":3.9032,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false},{"id":"image-video","weight":0.597,"sharePct":0.7982,"shareLowPct":0.06655,"shareHighPct":2.06205,"stale":false,"offBasis":true,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"avatar-video","weight":0.075,"sharePct":0.1005,"shareLowPct":0,"shareHighPct":1.0054,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"robotics-science","weight":0.033,"sharePct":0.1284,"shareLowPct":0,"shareHighPct":1.2837,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true}],"rising":null,"risingEligible":false,"risingComponents":{},"mc":{"medianRank":24,"rankP10":19,"rankP90":31,"pTop8":0,"pSelected":0,"stability":"borderline"},"baselineRevenueRank":null},{"entityId":"stability-ai","name":"Stability AI","reachPct":0.6962899,"importance":0.174072475,"importanceRank":28,"functions":[{"id":"speech-voice","weight":0.239,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"music","weight":0.056,"sharePct":0.5204,"shareLowPct":0,"shareHighPct":2.83628,"stale":true,"offBasis":true,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"image-video","weight":0.597,"sharePct":1.1175,"shareLowPct":0.162325,"shareHighPct":2.897575,"stale":true,"offBasis":true,"level":1,"years":0.25,"disagreement":false},{"id":"avatar-video","weight":0.075,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"robotics-science","weight":0.033,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false}],"rising":null,"risingEligible":false,"risingComponents":{},"mc":{"medianRank":28,"rankP10":21,"rankP90":36,"pTop8":0,"pSelected":0,"stability":"borderline"},"baselineRevenueRank":16},{"entityId":"mirage-captions","name":"Mirage (Captions)","reachPct":0.6786675,"importance":0.169666875,"importanceRank":29,"functions":[{"id":"speech-voice","weight":0.239,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"music","weight":0.056,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"image-video","weight":0.597,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"avatar-video","weight":0.075,"sharePct":9.0489,"shareLowPct":3.368175,"shareHighPct":20.359925,"stale":false,"offBasis":true,"level":1,"years":0.25,"disagreement":false},{"id":"robotics-science","weight":0.033,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false}],"rising":null,"risingEligible":false,"risingComponents":{},"mc":{"medianRank":28,"rankP10":21,"rankP90":35,"pTop8":0,"pSelected":0,"stability":"borderline"},"baselineRevenueRank":18},{"entityId":"hume-ai","name":"Hume AI","reachPct":0.5724767,"importance":0.143119175,"importanceRank":30,"functions":[{"id":"speech-voice","weight":0.239,"sharePct":2.3953,"shareLowPct":0,"shareHighPct":6.453925,"stale":false,"offBasis":true,"level":1,"years":0.25,"disagreement":false},{"id":"music","weight":0.056,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"image-video","weight":0.597,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"avatar-video","weight":0.075,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"robotics-science","weight":0.033,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false}],"rising":null,"risingEligible":false,"risingComponents":{},"mc":{"medianRank":31,"rankP10":22,"rankP90":38,"pTop8":0,"pSelected":0,"stability":"borderline"},"baselineRevenueRank":null},{"entityId":"chai-discovery","name":"Chai Discovery","reachPct":0.282414,"importance":0.141207,"importanceRank":31,"functions":[{"id":"speech-voice","weight":0.239,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"music","weight":0.056,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"image-video","weight":0.597,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"avatar-video","weight":0.075,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"robotics-science","weight":0.033,"sharePct":8.558,"shareLowPct":2.1395,"shareHighPct":34.2319,"stale":false,"offBasis":false,"level":2,"years":0.5,"disagreement":false}],"rising":null,"risingEligible":false,"risingComponents":{},"mc":{"medianRank":26,"rankP10":17,"rankP90":36,"pTop8":0.0001,"pSelected":0.0005,"stability":"borderline"},"baselineRevenueRank":null},{"entityId":"leonardo-ai","name":"Leonardo.Ai","reachPct":0.4765254,"importance":0.11913135,"importanceRank":32,"functions":[{"id":"speech-voice","weight":0.239,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"music","weight":0.056,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"image-video","weight":0.597,"sharePct":0.7982,"shareLowPct":0.2661,"shareHighPct":2.1286,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"avatar-video","weight":0.075,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"robotics-science","weight":0.033,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false}],"rising":null,"risingEligible":false,"risingComponents":{},"mc":{"medianRank":33,"rankP10":25,"rankP90":37,"pTop8":0,"pSelected":0,"stability":"borderline"},"baselineRevenueRank":null},{"entityId":"lightricks","name":"Lightricks (LTX)","reachPct":0.4765254,"importance":0.11913135,"importanceRank":33,"functions":[{"id":"speech-voice","weight":0.239,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"music","weight":0.056,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"image-video","weight":0.597,"sharePct":0.7982,"shareLowPct":0.3991,"shareHighPct":2.3947,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"avatar-video","weight":0.075,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"robotics-science","weight":0.033,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false}],"rising":null,"risingEligible":false,"risingComponents":{},"mc":{"medianRank":31,"rankP10":23,"rankP90":36,"pTop8":0,"pSelected":0,"stability":"borderline"},"baselineRevenueRank":null},{"entityId":"krea","name":"Krea","reachPct":0.4117068,"importance":0.1029267,"importanceRank":34,"functions":[{"id":"speech-voice","weight":0.239,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"music","weight":0.056,"sharePct":0.2604,"shareLowPct":0,"shareHighPct":1.3021,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"image-video","weight":0.597,"sharePct":0.6652,"shareLowPct":0.3725,"shareHighPct":1.8627,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"avatar-video","weight":0.075,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"robotics-science","weight":0.033,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false}],"rising":null,"risingEligible":false,"risingComponents":{},"mc":{"medianRank":33,"rankP10":26,"rankP90":37,"pTop8":0,"pSelected":0,"stability":"borderline"},"baselineRevenueRank":null},{"entityId":"luma-ai","name":"Luma AI","reachPct":0.3971244,"importance":0.0992811,"importanceRank":35,"functions":[{"id":"speech-voice","weight":0.239,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"music","weight":0.056,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"image-video","weight":0.597,"sharePct":0.6652,"shareLowPct":0.2129,"shareHighPct":3.1929,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"avatar-video","weight":0.075,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"robotics-science","weight":0.033,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false}],"rising":null,"risingEligible":false,"risingComponents":{},"mc":{"medianRank":30,"rankP10":20,"rankP90":37,"pTop8":0,"pSelected":0,"stability":"borderline"},"baselineRevenueRank":20},{"entityId":"shengshu","name":"Shengshu (Vidu)","reachPct":0.3971244,"importance":0.0992811,"importanceRank":36,"functions":[{"id":"speech-voice","weight":0.239,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"music","weight":0.056,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"image-video","weight":0.597,"sharePct":0.6652,"shareLowPct":0.2661,"shareHighPct":1.3305,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"avatar-video","weight":0.075,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"robotics-science","weight":0.033,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false}],"rising":null,"risingEligible":false,"risingComponents":{},"mc":{"medianRank":36,"rankP10":31,"rankP90":38,"pTop8":0,"pSelected":0,"stability":"borderline"},"baselineRevenueRank":19},{"entityId":"cartesia","name":"Cartesia","reachPct":0.3816352,"importance":0.0954088,"importanceRank":37,"functions":[{"id":"speech-voice","weight":0.239,"sharePct":1.5968,"shareLowPct":0.4258,"shareHighPct":4.7905,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false},{"id":"music","weight":0.056,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"image-video","weight":0.597,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"avatar-video","weight":0.075,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"robotics-science","weight":0.033,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false}],"rising":null,"risingEligible":false,"risingComponents":{},"mc":{"medianRank":34,"rankP10":26,"rankP90":38,"pTop8":0,"pSelected":0,"stability":"borderline"},"baselineRevenueRank":null},{"entityId":"decart","name":"Decart","reachPct":0.3177234,"importance":0.07943085,"importanceRank":38,"functions":[{"id":"speech-voice","weight":0.239,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"music","weight":0.056,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"image-video","weight":0.597,"sharePct":0.5322,"shareLowPct":0.133,"shareHighPct":2.1286,"stale":false,"offBasis":false,"level":1,"years":0.25,"disagreement":false,"defaultLevel":true},{"id":"avatar-video","weight":0.075,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false},{"id":"robotics-science","weight":0.033,"sharePct":0,"shareLowPct":0,"shareHighPct":0,"stale":false,"offBasis":false,"level":null,"years":null,"disagreement":false}],"rising":null,"risingEligible":false,"risingComponents":{},"mc":{"medianRank":34,"rankP10":26,"rankP90":38,"pTop8":0,"pSelected":0,"stability":"borderline"},"baselineRevenueRank":null}],"selection":{"importanceIds":["bytedance-seed","elevenlabs","suno","google-deepmind","kling-ai","midjourney","adobe-firefly","openai"],"risingIds":["skild-ai","heygen"],"filledFromImportance":[]},"risingPools":{"growth":7,"shareChange":7,"committed":1},"weightSensitivity":[{"functionId":"speech-voice","end":"low","entered":[],"left":[]},{"functionId":"speech-voice","end":"high","entered":[],"left":[]},{"functionId":"music","end":"low","entered":[],"left":[]},{"functionId":"music","end":"high","entered":[],"left":[]},{"functionId":"image-video","end":"low","entered":["synthesia"],"left":["adobe-firefly"]},{"functionId":"image-video","end":"high","entered":[],"left":[]},{"functionId":"avatar-video","end":"low","entered":[],"left":[]},{"functionId":"avatar-video","end":"high","entered":["synthesia"],"left":["openai"]},{"functionId":"robotics-science","end":"low","entered":[],"left":[]},{"functionId":"robotics-science","end":"high","entered":[],"left":[]}],"validation":[],"baseline":{"spearman":0.942857142857,"baselineTop10NotSelected":["soundhound","runway","synthesia","deepgram"],"selectedNotInBaselineTop10":["bytedance-seed","google-deepmind","openai","skild-ai"]},"raterAgreement":{"within":0.95,"weightedKappa":0.9,"disagreements":[]}},"inputs":{"bytedance-seed":{"schema":"5lc-index-inputs/1","key":"models-specialist-models","entityId":"bytedance-seed","name":"ByteDance Seed (Seedance, Seedream)","kind":"unit","parentId":"bytedance","commercialActivityVerified":true,"shares":{"speech-voice":{"low":0.5323,"central":2.1291,"high":7.9842,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Estimate for October 2026","periodEnd":"2026-10-22","source":{"url":"https://yicaiglobal.com/news/bytedances-volcano-engine-starts-price-war-in-chinas-enterprise-llms-sector","title":"ByteDance's Volcano Engine Starts Price War in China's Enterprise LLMs Sector","publisher":"Yicai Global","publishedDate":"2024-05-16","accessedDate":"2026-10-22","tier":"T5","locator":"Volcano Engine debuted nine Doubao models including pro and lite speech recognition, speech synthesis and text-to-picture models"},"note":"5LayerCake estimate (T5). Volcano Engine sells Doubao speech recognition and synthesis models (Yicai, May 2024); no revenue is disclosed. Allowance of about 2% of the speech part, range 0.5-8%. Share = the central US$ figure ÷ the part total of US$1,879M, which is the register candidates' central figures (US$1,779M) plus an outside-register allowance of US$100M for other speech developers not in the register (for example Gladia, Soniox, Fish Audio, Murf, WellSaid, Rev); the card's own estimate for this part is US$1,050M-1,440M-2,280M (low-central-high). Low and high shares hold the denominator at its central."},"music":{"low":0.5204,"central":2.0817,"high":7.8064,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Estimate for October 2026","periodEnd":"2026-10-22","source":{"url":"https://yicaiglobal.com/news/bytedances-volcano-engine-starts-price-war-in-chinas-enterprise-llms-sector","title":"ByteDance's Volcano Engine Starts Price War in China's Enterprise LLMs Sector","publisher":"Yicai Global","publishedDate":"2024-05-16","accessedDate":"2026-10-22","tier":"T5","locator":"Volcano Engine debuted nine Doubao models including pro and lite speech recognition, speech synthesis and text-to-picture models"},"note":"5LayerCake estimate (T5). No music-model product or revenue is stated in the cited page; the Doubao speech models are the closest product. Allowance for undisclosed revenue, not a measured figure. Share = the central US$ figure ÷ the part total of US$384M, which is the register candidates' central figures (US$364M) plus an outside-register allowance of US$20M for other music-model developers not in the register (for example Producer.ai, AIVA, Soundraw); the card's own estimate for this part is US$315M-336M-500M (low-central-high). Low and high shares hold the denominator at its central."},"image-video":{"low":9.4253,"central":23.5632,"high":42.4137,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Monthly revenue × 12, estimate for mid-2026","periodEnd":"2026-06-30","source":{"url":"https://eu.36kr.com/en/p/3869247060677636","title":"ByteDance AI, Sprinting Amid Anxiety","publisher":"36Kr (reprint of China Entrepreneur Magazine)","publishedDate":"2026-06-26","accessedDate":"2026-10-22","tier":"T5","locator":"Volcengine president Tan Dai: Seedance revenue figures circulating outside the company are all wrong and on the high side"},"note":"5LayerCake estimate (T5). Insiders told 36Kr in June 2026 that Seedance 2.0 alone can earn more than RMB1bn a month for Volcano Engine (KrASIA, 5 June 2026, https://kr-asia.com/bytedance-raises-volcano-engines-maas-revenue-target-on-seedance-2-0-growth); Volcano Engine raised its 2026 model-as-a-service revenue target to RMB15bn, against about RMB1.5bn in 2025 (same article). Volcengine president Tan Dai said that the Seedance revenue figures circulating outside the company are 'all wrong and on the high side' (36Kr, 26 June 2026). A whole-year target of RMB15bn for every model, mostly language-model tokens, is RMB1.25bn a month on average, so even RMB0.9bn a month for video alone is generous. Method: RMB0.2 / 0.5 / 0.9bn a month × 12 ÷ 6.7758 (June 2026 average rate, Federal Reserve H.10 monthly averages, FRED series EXCHUS, https://fred.stlouisfed.org/data/EXCHUS.txt, last updated 1 October 2026) = US$354 / 886 / 1,594M. Seedance and Seedream APIs and the Dreamina/Jimeng apps; Doubao-app features count in Applications. Vercel's August 2026 gateway data ranks Seedance first by videos and video spend (https://vercel.com/blog/ai-gateway-production-index-september-2026). Unit figure, not ByteDance's total. Share = the central US$ figure ÷ the part total of US$3,758M. That is the screening total (the register candidates' central figures of US$3,608M plus an outside-register allowance of US$150M for other developers not in the register), held fixed so every company in this sublayer is measured against one total; the card's own estimate for this part is US$2,140M-3,590M-5,660M (low-central-high). Low and high shares hold the denominator at its central."},"avatar-video":{"low":0,"central":0.4022,"high":3.0163,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Estimate for October 2026","periodEnd":"2026-10-22","source":{"url":"https://eu.36kr.com/en/p/3869247060677636","title":"ByteDance AI, Sprinting Amid Anxiety","publisher":"36Kr (reprint of China Entrepreneur Magazine)","publishedDate":"2026-06-26","accessedDate":"2026-10-22","tier":"T5","locator":"Volcengine president Tan Dai: Seedance revenue figures circulating outside the company are all wrong and on the high side"},"note":"5LayerCake estimate (T5). No avatar-video product or revenue is stated in the cited article. Allowance for undisclosed revenue, not a measured figure. Share = the central US$ figure ÷ the part total of US$497M, which is the register candidates' central figures (US$467M) plus an outside-register allowance of US$30M for other avatar-video developers not in the register (for example Akool, Argil, Creatify); the card's own estimate for this part is US$400M-450M-550M (low-central-high). Low and high shares hold the denominator at its central."},"robotics-science":{"participates":false,"evidence":{"url":"https://eu.36kr.com/en/p/3869247060677636","title":"ByteDance AI, Sprinting Amid Anxiety","publisher":"36Kr (reprint of China Entrepreneur Magazine)","publishedDate":"2026-06-26","accessedDate":"2026-10-22","tier":"T3","locator":"36Kr describes ByteDance's AI business (Doubao models, Seedance, Volcano Engine); it shows no robot-control or science-model product sold."}}},"screening":{"speech-voice":{"levelLow":1,"levelHigh":2},"music":{"levelLow":1,"levelHigh":2},"image-video":{"levelLow":1,"levelHigh":3},"avatar-video":{"levelLow":1,"levelHigh":2}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"664e7d2886629c589472711f356fccc02560e5c11fdb80fd97dac1bb45691ba6"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"664e7d2886629c589472711f356fccc02560e5c11fdb80fd97dac1bb45691ba6"}}},"elevenlabs":{"schema":"5lc-index-inputs/1","key":"models-specialist-models","entityId":"elevenlabs","name":"ElevenLabs","kind":"private","commercialActivityVerified":true,"shares":{"speech-voice":{"low":31.937,"central":31.937,"high":31.937,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"ARR, September 2026","periodEnd":"2026-09-24","source":{"url":"https://techcrunch.com/2026/09/24/twenty-minutes-with-the-ceo-of-elevenlabs-now-reportedly-valued-at-22-billion/","title":"ElevenLabs' CEO on margins, IPO timing, and telling customers they're talking to a bot","publisher":"TechCrunch","publishedDate":"2026-09-24","accessedDate":"2026-10-22","tier":"T2","locator":"CEO Mati Staniszewski: \"We are now $600 million in ARR\"; 55%+ classic enterprise"},"note":"The company says it is at $600m ARR (CEO interview, 24 Sept 2026); 55%+ from classic enterprise. The figure covers voice agents, dubbing and music, so music is also counted below at a small allowance and speech may be slightly overstated (inside the T2 band). Share = the central US$ figure ÷ the part total of US$1,879M, which is the register candidates' central figures (US$1,779M) plus an outside-register allowance of US$100M for other speech developers not in the register (for example Gladia, Soniox, Fish Audio, Murf, WellSaid, Rev); the card's own estimate for this part is US$1,050M-1,440M-2,280M (low-central-high). Low and high shares hold the denominator at its central."},"music":{"low":1.3011,"central":3.9032,"high":10.4085,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"ARR share, September 2026 (estimate)","periodEnd":"2026-09-24","source":{"url":"https://techcrunch.com/2026/09/24/twenty-minutes-with-the-ceo-of-elevenlabs-now-reportedly-valued-at-22-billion/","title":"ElevenLabs' CEO on margins, IPO timing, and telling customers they're talking to a bot","publisher":"TechCrunch","publishedDate":"2026-09-24","accessedDate":"2026-10-22","tier":"T5","locator":"CEO Mati Staniszewski: \"We are now $600 million in ARR\"; 55%+ classic enterprise"},"note":"5LayerCake estimate (T5). The article lists music among ElevenLabs' creator products (the writer's description, no figure). Allowance of 1-7% of the $600m ARR, which sits inside the speech figure. Share = the central US$ figure ÷ the part total of US$384M, which is the register candidates' central figures (US$364M) plus an outside-register allowance of US$20M for other music-model developers not in the register (for example Producer.ai, AIVA, Soundraw); the card's own estimate for this part is US$315M-336M-500M (low-central-high). Low and high shares hold the denominator at its central."},"image-video":{"participates":false,"evidence":{"url":"https://techcrunch.com/2026/09/24/twenty-minutes-with-the-ceo-of-elevenlabs-now-reportedly-valued-at-22-billion/","title":"ElevenLabs' CEO on margins, IPO timing, and telling customers they're talking to a bot","publisher":"TechCrunch","publishedDate":"2026-09-24","accessedDate":"2026-10-22","tier":"T3","locator":"The interview describes ElevenLabs' products as speech and voice agents, dubbing and music; it names no image, video, avatar or robot models."}},"avatar-video":{"participates":false,"evidence":{"url":"https://techcrunch.com/2026/09/24/twenty-minutes-with-the-ceo-of-elevenlabs-now-reportedly-valued-at-22-billion/","title":"ElevenLabs' CEO on margins, IPO timing, and telling customers they're talking to a bot","publisher":"TechCrunch","publishedDate":"2026-09-24","accessedDate":"2026-10-22","tier":"T3","locator":"The interview describes ElevenLabs' products as speech and voice agents, dubbing and music; it names no image, video, avatar or robot models."}},"robotics-science":{"participates":false,"evidence":{"url":"https://techcrunch.com/2026/09/24/twenty-minutes-with-the-ceo-of-elevenlabs-now-reportedly-valued-at-22-billion/","title":"ElevenLabs' CEO on margins, IPO timing, and telling customers they're talking to a bot","publisher":"TechCrunch","publishedDate":"2026-09-24","accessedDate":"2026-10-22","tier":"T3","locator":"The interview describes ElevenLabs' products as speech and voice agents, dubbing and music; it names no image, video, avatar or robot models."}}},"priorShares":{"speech-voice":{"low":23.8425,"central":23.8425,"high":23.8425,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"ARR, September 2025 (surpassed earlier in 2025)","periodEnd":"2025-09-08","source":{"url":"https://tech.eu/2025/09/08/elevenlabs-confirms-employee-share-sale-at-6-6bn-valuation/","title":"ElevenLabs confirms employee share sale at $6.6BN valuation, double valuation of nine months ago","publisher":"Tech.eu","publishedDate":"2025-09-08","tier":"T2","locator":"Quotes CEO Mati Staniszewski's LinkedIn post: \"Earlier this year, we surpassed $200 million in ARR and we expect to top $300m by year end\"","accessedDate":"2026-10-22"},"note":"The company says (CEO Mati Staniszewski, LinkedIn post of September 2025, quoted by Tech.eu on 8 September 2025) that it had surpassed US$200M of ARR earlier in 2025. 'Surpassed' makes US$200M a floor. Its ARR includes its music and other products, as does the current US$600M. The trajectory is consistent with the CEO's later statement, reported on 13 January 2026 from a Bloomberg interview, that ARR crossed US$330M in 2025, after 20 months to US$100M and 10 months to US$200M. Prior share = the prior US$ figure ÷ the prior part total of US$839M, which is the current part total of US$1,879M ÷ 2.24, the growth of this part over the year: pooled from the four speech companies with a sourced current and prior figure (ElevenLabs US$600M ÷ US$200M, AssemblyAI US$76M ÷ US$26M, SoundHound's counted share US$117.7M ÷ US$94.2M, iFLYTEK's counted share US$83.5M ÷ US$70.8M), US$877.2M ÷ US$391.0M = 2.24. Companies without a sourced prior are assumed to have grown at that rate. Low and high hold the denominator at its central."},"music":{"low":1.3011,"central":3.9032,"high":10.4085,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Held at the current share (prior period 2025-09-24)","periodEnd":"2025-09-24","source":{"url":"https://techcrunch.com/2026/09/24/twenty-minutes-with-the-ceo-of-elevenlabs-now-reportedly-valued-at-22-billion/","title":"ElevenLabs' CEO on margins, IPO timing, and telling customers they're talking to a bot","publisher":"TechCrunch","publishedDate":"2026-09-24","accessedDate":"2026-10-22","tier":"T5","locator":"CEO Mati Staniszewski: \"We are now $600 million in ARR\"; 55%+ classic enterprise"},"note":"5LayerCake assumption (T5): no figure for the earlier period exists for this undisclosed allowance, so the share is held at the current share (the company is assumed to have grown with this part of the market). ElevenLabs launched its music product in August 2025, so the share may have been lower a year ago; no figure exists."}},"activityGrowth":{"current":{"low":600000000,"central":600000000,"high":600000000,"unit":"USD","basis":"Annual recurring revenue (ARR) of ElevenLabs, US dollars (company statements)","period":"ARR, September 2026","periodEnd":"2026-09-24","source":{"url":"https://techcrunch.com/2026/09/24/twenty-minutes-with-the-ceo-of-elevenlabs-now-reportedly-valued-at-22-billion/","title":"ElevenLabs' CEO on margins, IPO timing, and telling customers they're talking to a bot","publisher":"TechCrunch","publishedDate":"2026-09-24","accessedDate":"2026-10-22","tier":"T2","locator":"CEO Mati Staniszewski: \"We are now $600 million in ARR\"; 55%+ classic enterprise"},"note":"The company says (CEO Mati Staniszewski, TechCrunch interview, 24 September 2026): 'We are now $600 million in ARR'."},"prior":{"low":200000000,"central":200000000,"high":200000000,"unit":"USD","basis":"Annual recurring revenue (ARR) of ElevenLabs, US dollars (company statements)","period":"ARR, September 2025 (surpassed earlier in 2025)","periodEnd":"2025-09-08","source":{"url":"https://tech.eu/2025/09/08/elevenlabs-confirms-employee-share-sale-at-6-6bn-valuation/","title":"ElevenLabs confirms employee share sale at $6.6BN valuation, double valuation of nine months ago","publisher":"Tech.eu","publishedDate":"2025-09-08","tier":"T2","locator":"Quotes CEO Mati Staniszewski's LinkedIn post: \"Earlier this year, we surpassed $200 million in ARR and we expect to top $300m by year end\"","accessedDate":"2026-10-22"},"note":"'Surpassed' US$200M, so a floor; the growth ratio 600 ÷ 200 = 3.0 is therefore a ceiling for the true ratio."}},"baselineRevenueUsd":{"low":600000000,"central":600000000,"high":600000000,"unit":"USD","basis":"Annual recurring revenue (ARR) of ElevenLabs, US dollars (company statements)","period":"ARR, September 2026","periodEnd":"2026-09-24","source":{"url":"https://techcrunch.com/2026/09/24/twenty-minutes-with-the-ceo-of-elevenlabs-now-reportedly-valued-at-22-billion/","title":"ElevenLabs' CEO on margins, IPO timing, and telling customers they're talking to a bot","publisher":"TechCrunch","publishedDate":"2026-09-24","accessedDate":"2026-10-22","tier":"T2","locator":"CEO Mati Staniszewski: \"We are now $600 million in ARR\"; 55%+ classic enterprise"},"note":"ARR is a run-rate, not recognised revenue. ElevenLabs' own figure covers all its products (voices, agents, dubbing, music); 55% or more is classic enterprise."},"screening":{"speech-voice":{"levelLow":1,"levelHigh":3},"music":{"levelLow":1,"levelHigh":2}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"1a1fdc2a5d3e6cfd8fc2f7f457fa88834d3bdfefa2ea2f5def02b993d017b646"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"1a1fdc2a5d3e6cfd8fc2f7f457fa88834d3bdfefa2ea2f5def02b993d017b646"}}},"suno":{"schema":"5lc-index-inputs/1","key":"models-specialist-models","entityId":"suno","name":"Suno","kind":"private","commercialActivityVerified":true,"shares":{"speech-voice":{"participates":false,"evidence":{"url":"https://techcrunch.com/2026/02/27/ai-music-generator-suno-hits-2-million-paid-subscribers-and-300m-in-annual-recurring-revenue/","title":"AI music generator Suno hits 2 million paid subscribers and $300M in annual recurring revenue","publisher":"TechCrunch","publishedDate":"2026-02-27","accessedDate":"2026-10-22","tier":"T3","locator":"TechCrunch describes Suno as an AI music generator sold by subscription; no speech, image, video, avatar or robot models."}},"music":{"low":78.064,"central":78.064,"high":78.064,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"ARR, February 2026","periodEnd":"2026-02-27","source":{"url":"https://techcrunch.com/2026/02/27/ai-music-generator-suno-hits-2-million-paid-subscribers-and-300m-in-annual-recurring-revenue/","title":"AI music generator Suno hits 2 million paid subscribers and $300M in annual recurring revenue","publisher":"TechCrunch","publishedDate":"2026-02-27","accessedDate":"2026-10-22","tier":"T2","locator":"Suno's CEO on LinkedIn: 2 million paid subscribers and $300 million ARR"},"note":"The company says $300M ARR and 2 million paid subscribers (CEO on LinkedIn, reported 27 Feb 2026); $200M annual revenue three months earlier (WSJ, per TechCrunch). Newer than the 12-month stale limit. Share = the central US$ figure ÷ the part total of US$384M, which is the register candidates' central figures (US$364M) plus an outside-register allowance of US$20M for other music-model developers not in the register (for example Producer.ai, AIVA, Soundraw); the card's own estimate for this part is US$315M-336M-500M (low-central-high). Low and high shares hold the denominator at its central."},"image-video":{"participates":false,"evidence":{"url":"https://techcrunch.com/2026/02/27/ai-music-generator-suno-hits-2-million-paid-subscribers-and-300m-in-annual-recurring-revenue/","title":"AI music generator Suno hits 2 million paid subscribers and $300M in annual recurring revenue","publisher":"TechCrunch","publishedDate":"2026-02-27","accessedDate":"2026-10-22","tier":"T3","locator":"TechCrunch describes Suno as an AI music generator sold by subscription; no speech, image, video, avatar or robot models."}},"avatar-video":{"participates":false,"evidence":{"url":"https://techcrunch.com/2026/02/27/ai-music-generator-suno-hits-2-million-paid-subscribers-and-300m-in-annual-recurring-revenue/","title":"AI music generator Suno hits 2 million paid subscribers and $300M in annual recurring revenue","publisher":"TechCrunch","publishedDate":"2026-02-27","accessedDate":"2026-10-22","tier":"T3","locator":"TechCrunch describes Suno as an AI music generator sold by subscription; no speech, image, video, avatar or robot models."}},"robotics-science":{"participates":false,"evidence":{"url":"https://techcrunch.com/2026/02/27/ai-music-generator-suno-hits-2-million-paid-subscribers-and-300m-in-annual-recurring-revenue/","title":"AI music generator Suno hits 2 million paid subscribers and $300M in annual recurring revenue","publisher":"TechCrunch","publishedDate":"2026-02-27","accessedDate":"2026-10-22","tier":"T3","locator":"TechCrunch describes Suno as an AI music generator sold by subscription; no speech, image, video, avatar or robot models."}}},"baselineRevenueUsd":{"low":300000000,"central":300000000,"high":300000000,"unit":"USD","basis":"Annual recurring revenue (ARR) of Suno, US dollars (company statement)","period":"ARR, February 2026","periodEnd":"2026-02-27","source":{"url":"https://techcrunch.com/2026/02/27/ai-music-generator-suno-hits-2-million-paid-subscribers-and-300m-in-annual-recurring-revenue/","title":"AI music generator Suno hits 2 million paid subscribers and $300M in annual recurring revenue","publisher":"TechCrunch","publishedDate":"2026-02-27","accessedDate":"2026-10-22","tier":"T2","locator":"Suno's CEO on LinkedIn: 2 million paid subscribers and $300 million ARR"},"note":"The company says US$300M ARR and 2 million paid subscribers (CEO Mikey Shulman, LinkedIn post, reported by TechCrunch on 27 February 2026). The Series D of June 2026 (US$5.4B valuation) gave no newer revenue figure."},"screening":{"music":{"levelLow":1,"levelHigh":3}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"5a437f60a329993e18f9f500a0ddf2e4edc70242fe5fc2c1dc0661021e9ac46c"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"5a437f60a329993e18f9f500a0ddf2e4edc70242fe5fc2c1dc0661021e9ac46c"}}},"google-deepmind":{"schema":"5lc-index-inputs/1","key":"models-specialist-models","entityId":"google-deepmind","name":"Google DeepMind","kind":"unit","parentId":"alphabet","commercialActivityVerified":true,"shares":{"speech-voice":{"low":3.1932,"central":9.0474,"high":21.2879,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Estimate for October 2026","periodEnd":"2026-10-22","source":{"url":"https://ai.google.dev/gemini-api/docs/pricing","title":"Gemini Developer API pricing","publisher":"Google","publishedDate":"undated","tier":"T5","locator":"Gemini TTS models are on the pay-as-you-go price list (for example Gemini 3.1 Flash TTS Preview at US$1.00 per million input text tokens and US$20.00 per million output audio tokens)","accessedDate":"2026-10-22"},"note":"5LayerCake estimate (T5). Google sells Gemini TTS through the Gemini API (price list opened 22 October 2026), and Cloud Text-to-Speech, Chirp and speech-to-text on Google Cloud, and discloses no revenue or minutes. The card's allowance for the four large companies' speech services is US$300-600-1,200M (0.5-1-2 times ElevenLabs); Google, Microsoft, Amazon and OpenAI are given 170 / 170 / 140 / 80 of the central. Real-time voice in general assistants counts in Frontier labs. Share = the central US$ figure ÷ the part total of US$1,879M. That is the screening total (the register candidates' central figures of US$1,779M plus an outside-register allowance of US$100M for other developers not in the register), held fixed so every company in this sublayer is measured against one total; the card's own estimate for this part is US$1,050M-1,440M-2,280M (low-central-high). Low and high shares hold the denominator at its central."},"music":{"low":0.5208,"central":2.0833,"high":7.8125,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Estimate for October 2026","periodEnd":"2026-10-22","source":{"url":"https://ai.google.dev/gemini-api/docs/pricing","title":"Gemini Developer API pricing","publisher":"Google","publishedDate":"undated","tier":"T5","locator":"Lyria 3.5 full songs at US$0.08 a song and Lyria 3 Clip Preview at US$0.04 a 30-second clip on the Gemini API price list","accessedDate":"2026-10-22"},"note":"5LayerCake estimate (T5). Google sells Lyria music models through the Gemini API (US$0.08 a song, price list opened 22 October 2026) and Google Cloud (Lyria 3 alongside Lyria 2); no revenue is disclosed. Allowance of about 2% of the music part: central US$8M, high US$30M. Share = the central US$ figure ÷ the part total of US$384M. That is the screening total (the register candidates' central figures of US$364M plus an outside-register allowance of US$20M for other developers not in the register), held fixed so every company in this sublayer is measured against one total; the card's own estimate for this part is US$315M-336M-500M (low-central-high). Low and high shares hold the denominator at its central."},"image-video":{"low":3.1929,"central":10.6431,"high":23.947,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Estimate for October 2026","periodEnd":"2026-09-17","source":{"url":"https://vercel.com/blog/ai-gateway-production-index-september-2026","title":"Open-weight models take 56% of token volume, Astra doubles Fable 5.1 spend (AI Gateway Production Index, September 2026)","publisher":"Vercel","publishedDate":"2026-09-17","accessedDate":"2026-10-22","tier":"T5","locator":"August 2026 data: Nano Banana 50% of image spend vs GPT Image 44%; Seedance first in videos and video spend; Veo 20% of video spend; Grok Imagine 19% of videos"},"note":"5LayerCake estimate (T5). Google (Veo, Imagen and Nano Banana via the Gemini API and Vertex) does not disclose revenue. The card's allowance for Google, OpenAI, Amazon and Microsoft image and video is US$300-800-1,600M, split by Vercel's August 2026 gateway data: Nano Banana 50% of image spend (GPT Image 44%) and Veo 20% of video spend (Seedance first). Google about 50% = US$400M, range 120-900M. App use (Gemini app, Flow) counts in Applications. Unit figure, not Alphabet's total. Share = the central US$ figure ÷ the part total of US$3,758M, which is the register candidates' central figures (US$3,608M) plus an outside-register allowance of US$150M for other image and video developers not in the register (for example Higgsfield, Freepik, Tripo, Genmo); the card's own estimate for this part is US$2,140M-3,590M-5,660M (low-central-high). Low and high shares hold the denominator at its central."},"avatar-video":{"low":0,"central":0.2011,"high":2.0109,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Estimate for October 2026","periodEnd":"2026-10-22","source":{"url":"https://vercel.com/blog/ai-gateway-production-index-september-2026","title":"Open-weight models take 56% of token volume, Astra doubles Fable 5.1 spend (AI Gateway Production Index, September 2026)","publisher":"Vercel","publishedDate":"2026-09-17","accessedDate":"2026-10-22","tier":"T5","locator":"August 2026 data: Nano Banana 50% of image spend vs GPT Image 44%; Seedance first in videos and video spend; Veo 20% of video spend; Grok Imagine 19% of videos"},"note":"5LayerCake estimate (T5). No stand-alone avatar-video product was found for Google DeepMind in the cited page; avatar features in Workspace apps count in Applications. Allowance for undisclosed revenue, not a measured figure. Share = the central US$ figure ÷ the part total of US$497M, which is the register candidates' central figures (US$467M) plus an outside-register allowance of US$30M for other avatar-video developers not in the register (for example Akool, Argil, Creatify); the card's own estimate for this part is US$400M-450M-550M (low-central-high). Low and high shares hold the denominator at its central."},"robotics-science":{"low":0,"central":0.8547,"high":4.2735,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Estimate for October 2026","periodEnd":"2026-10-22","source":{"url":"https://ai.google.dev/gemini-api/docs/pricing","title":"Gemini Developer API pricing","publisher":"Google","publishedDate":"undated","tier":"T5","locator":"Gemini Robotics ER 2 Preview at US$1.00 per million input tokens and US$5.00 per million output tokens through 31 December 2026 (both double from 1 January 2027)","accessedDate":"2026-10-22"},"note":"5LayerCake estimate (T5). Gemini Robotics ER 2 (preview, announced in the Gemini API changelog on 30 July 2026) is sold through the Gemini API at a stated price, so it can earn revenue; none is disclosed. Allowance only: central US$2M, high US$10M. Isomorphic Labs' partnership payments are excluded under the controller note on science models (Isomorphic is consolidated into this entry). Share = the central US$ figure ÷ the part total of US$234M. That is the screening total (the register candidates' central figures of US$219M plus an outside-register allowance of US$15M for other developers not in the register), held fixed so every company in this sublayer is measured against one total; the card's own estimate for this part is US$140M-200M-350M (low-central-high). Low and high shares hold the denominator at its central."}},"screening":{"speech-voice":{"levelLow":1,"levelHigh":3},"music":{"levelLow":1,"levelHigh":2},"image-video":{"levelLow":1,"levelHigh":3},"avatar-video":{"levelLow":1,"levelHigh":2},"robotics-science":{"levelLow":1,"levelHigh":3}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"c6276ad0b8dd7286f85d4f0ac2a386ecc55614c50876388a4d40ca973c46aee6"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"c6276ad0b8dd7286f85d4f0ac2a386ecc55614c50876388a4d40ca973c46aee6"}}},"kling-ai":{"schema":"5lc-index-inputs/1","key":"models-specialist-models","entityId":"kling-ai","name":"Kling AI","kind":"unit","parentId":"kuaishou","commercialActivityVerified":true,"shares":{"speech-voice":{"participates":false,"evidence":{"url":"https://www1.hkexnews.hk/listedco/listconews/sehk/2026/0819/2026081900314.pdf","title":"Kuaishou Technology results announcement for the three and six months ended June 30, 2026","publisher":"Kuaishou Technology (HKEXnews)","publishedDate":"2026-08-19","accessedDate":"2026-10-22","tier":"T1","locator":"Kuaishou's results announcement describes Kling AI as an AI video generation business; it shows no speech, music, avatar or robot model products for Kling."}},"music":{"participates":false,"evidence":{"url":"https://www1.hkexnews.hk/listedco/listconews/sehk/2026/0819/2026081900314.pdf","title":"Kuaishou Technology results announcement for the three and six months ended June 30, 2026","publisher":"Kuaishou Technology (HKEXnews)","publishedDate":"2026-08-19","accessedDate":"2026-10-22","tier":"T1","locator":"Kuaishou's results announcement describes Kling AI as an AI video generation business; it shows no speech, music, avatar or robot model products for Kling."}},"image-video":{"low":13.2968,"central":13.2968,"high":13.2968,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Q2 2026 revenue × 4, converted at RMB 6.804/US$ (Q2 2026 average)","periodEnd":"2026-06-30","source":{"url":"https://www1.hkexnews.hk/listedco/listconews/sehk/2026/0819/2026081900314.pdf","title":"Kuaishou Technology results announcement for the three and six months ended June 30, 2026","publisher":"Kuaishou Technology (HKEXnews)","publishedDate":"2026-08-19","accessedDate":"2026-10-22","tier":"T1","locator":"p. 5, AI business: Kling AI generated revenue of over RMB850 million in the second quarter of 2026, up more than 200% year over year"},"note":"Filing (T1). Kuaishou reports Kling AI revenue of over RMB850M in Q2 2026 (p. 5), up more than 200% year over year: RMB850M × 4 ÷ 6.804 = US$499.7M. The rate is the Q2 2026 average of the Federal Reserve H.10 monthly averages (Federal Reserve H.10 monthly averages, FRED series EXCHUS, https://fred.stlouisfed.org/data/EXCHUS.txt, last updated 1 October 2026): (6.8371 + 6.7996 + 6.7758) ÷ 3. 'Over' means the true figure is at least this. Unit figure, never Kuaishou's total. Share = the central US$ figure ÷ the part total of US$3,758M. That is the screening total (the register candidates' central figures of US$3,608M plus an outside-register allowance of US$150M for other developers not in the register), held fixed so every company in this sublayer is measured against one total; the card's own estimate for this part is US$2,140M-3,590M-5,660M (low-central-high). Low and high shares hold the denominator at its central."},"avatar-video":{"participates":false,"evidence":{"url":"https://www1.hkexnews.hk/listedco/listconews/sehk/2026/0819/2026081900314.pdf","title":"Kuaishou Technology results announcement for the three and six months ended June 30, 2026","publisher":"Kuaishou Technology (HKEXnews)","publishedDate":"2026-08-19","accessedDate":"2026-10-22","tier":"T1","locator":"Kuaishou's results announcement describes Kling AI as an AI video generation business; it shows no speech, music, avatar or robot model products for Kling."}},"robotics-science":{"participates":false,"evidence":{"url":"https://www1.hkexnews.hk/listedco/listconews/sehk/2026/0819/2026081900314.pdf","title":"Kuaishou Technology results announcement for the three and six months ended June 30, 2026","publisher":"Kuaishou Technology (HKEXnews)","publishedDate":"2026-08-19","accessedDate":"2026-10-22","tier":"T1","locator":"Kuaishou's results announcement describes Kling AI as an AI video generation business; it shows no speech, music, avatar or robot model products for Kling."}}},"priorShares":{"image-video":{"low":8.9423,"central":8.9423,"high":8.9423,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Q2 2025 revenue × 4, converted at RMB 7.231/US$ (Q2 2025 average)","periodEnd":"2025-06-30","source":{"url":"https://www.prnewswire.com/apac/news-releases/kuaishou-technology-announces-second-quarter-and-interim-2025-unaudited-financial-results-and-declares-a-special-dividend-302535640.html","title":"Kuaishou Technology Announces Second Quarter and Interim 2025 Unaudited Financial Results and Declares a Special Dividend","publisher":"Kuaishou Technology (PR Newswire)","publishedDate":"2025-08-21","tier":"T2","locator":"\"In the second quarter of 2025, revenue from Kling AI (可靈AI) surpassed RMB250 million.\"","accessedDate":"2026-10-22"},"note":"The company says (results release of 21 August 2025, T2; the HKEX results announcement carries the same text) that Kling AI revenue surpassed RMB250M in Q2 2025: RMB250M × 4 ÷ 7.231 (Q2 2025 average of the H.10 monthly averages: (7.2968 + 7.2166 + 7.1804) ÷ 3) = US$138.3M. 'Surpassed' makes this a floor, as 'over' does for the current figure. Prior share = the prior US$ figure ÷ the prior part total of US$1,547M, which is the current part total of US$3,758M ÷ 2.43, the growth of this part over the year: pooled from every company in this sublayer's file set with a sourced pair, because the image-and-video pairs alone (Kling US$500M ÷ US$138M, Meshy at 50% US$50M ÷ US$7.5M, MiniMax's video share US$30M ÷ US$15M) cover only 15% of this part: the four speech pairs (ElevenLabs US$600M ÷ US$200M, AssemblyAI US$76M ÷ US$26M, SoundHound US$117.7M ÷ US$94.2M, iFLYTEK US$83.5M ÷ US$70.8M) and the two avatar pairs (HeyGen US$205M ÷ US$95M, Synthesia US$150M ÷ US$100M) plus these, US$1,812M ÷ US$747M = 2.43. Companies without a sourced prior are assumed to have grown at that rate. Low and high hold the denominator at its central."}},"activityGrowth":{"current":{"low":850000000,"central":850000000,"high":850000000,"unit":"CNY","basis":"Kling AI quarterly revenue, RMB (company-reported)","period":"Q2 2026 (calendar quarter)","periodEnd":"2026-06-30","source":{"url":"https://www1.hkexnews.hk/listedco/listconews/sehk/2026/0819/2026081900314.pdf","title":"Kuaishou Technology results announcement for the three and six months ended June 30, 2026","publisher":"Kuaishou Technology (HKEXnews)","publishedDate":"2026-08-19","accessedDate":"2026-10-22","tier":"T1","locator":"p. 5: Kling AI generated revenue of over RMB850 million in the second quarter of 2026"},"note":"'Over RMB850M': a floor."},"prior":{"low":250000000,"central":250000000,"high":250000000,"unit":"CNY","basis":"Kling AI quarterly revenue, RMB (company-reported)","period":"Q2 2025 (calendar quarter)","periodEnd":"2025-06-30","source":{"url":"https://www.prnewswire.com/apac/news-releases/kuaishou-technology-announces-second-quarter-and-interim-2025-unaudited-financial-results-and-declares-a-special-dividend-302535640.html","title":"Kuaishou Technology Announces Second Quarter and Interim 2025 Unaudited Financial Results and Declares a Special Dividend","publisher":"Kuaishou Technology (PR Newswire)","publishedDate":"2025-08-21","tier":"T2","locator":"\"In the second quarter of 2025, revenue from Kling AI (可靈AI) surpassed RMB250 million.\"","accessedDate":"2026-10-22"},"note":"'Surpassed RMB250M': a floor. 850 ÷ 250 = 3.4 times (the report itself says more than 200% growth)."}},"baselineRevenueUsd":{"low":499693815.0642988,"central":499693815.0642988,"high":499693815.0642988,"unit":"USD","basis":"Kling AI revenue, latest calendar quarter × 4, US dollars","period":"Q2 2026 revenue × 4","periodEnd":"2026-06-30","source":{"url":"https://www1.hkexnews.hk/listedco/listconews/sehk/2026/0819/2026081900314.pdf","title":"Kuaishou Technology results announcement for the three and six months ended June 30, 2026","publisher":"Kuaishou Technology (HKEXnews)","publishedDate":"2026-08-19","accessedDate":"2026-10-22","tier":"T1","locator":"p. 5: Kling AI generated revenue of over RMB850 million in the second quarter of 2026"},"note":"RMB850M × 4 ÷ 6.804 (Q2 2026 average rate). The segment is Kling AI only, never Kuaishou's total."},"screening":{"image-video":{"levelLow":1,"levelHigh":3}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"6c21a586479812e32becab3bf008b84c6c9e13c467a4a8ad4986d25c0a5a6682"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"6c21a586479812e32becab3bf008b84c6c9e13c467a4a8ad4986d25c0a5a6682"}}},"midjourney":{"schema":"5lc-index-inputs/1","key":"models-specialist-models","entityId":"midjourney","name":"Midjourney","kind":"private","commercialActivityVerified":true,"shares":{"speech-voice":{"participates":false,"evidence":{"url":"https://winbuzzer.com/2026/03/31/midjourney-revenue-above-200-million-hardware-push-xcxwbn/","title":"Midjourney Revenue Tops $200M as It Eyes Hardware","publisher":"WinBuzzer","publishedDate":"2026-03-31","accessedDate":"2026-10-22","tier":"T3","locator":"WinBuzzer describes Midjourney as an image (and video) generator sold by subscription; no speech, music, avatar or robot models."}},"music":{"participates":false,"evidence":{"url":"https://winbuzzer.com/2026/03/31/midjourney-revenue-above-200-million-hardware-push-xcxwbn/","title":"Midjourney Revenue Tops $200M as It Eyes Hardware","publisher":"WinBuzzer","publishedDate":"2026-03-31","accessedDate":"2026-10-22","tier":"T3","locator":"WinBuzzer describes Midjourney as an image (and video) generator sold by subscription; no speech, music, avatar or robot models."}},"image-video":{"low":7.983,"central":10.644,"high":14.6354,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Run-rate, estimate for 2026","periodEnd":"2026-03-29","source":{"url":"https://www.theinformation.com/briefings/midjourney-revenue-now-significantly-200-million","title":"Midjourney Revenue Now 'Significantly' Above $200 Million","publisher":"The Information","publishedDate":"undated","tier":"T5","locator":"Briefing headline by Abram Brown (the page shows no date and the body is behind a paywall); WinBuzzer dates the CEO interview to 29 March 2026","accessedDate":"2026-10-22"},"note":"5LayerCake estimate (T5). The company says its revenue is well above US$200M and still growing: The Information's briefing is headed 'Midjourney Revenue Now Significantly Above $200 Million' (body paywalled), and WinBuzzer's account of the same interview with CEO David Holz (29 March 2026; https://winbuzzer.com/2026/03/31/midjourney-revenue-above-200-million-hardware-push-xcxwbn/) says 2023 revenue 'significantly surpassed' US$200M and has continued growing. TIME's profile of Holz, as relayed in the same WinBuzzer article ('Per Time's profile of Holz, revenue pushed to $300 million by 2024'), puts revenue at US$300M by 2024. Method: a 2023 base of US$200-250M × 1.2-2.2 growth to 2026, central US$400M; low US$300M (the TIME figure for 2024); high US$550M. Data-broker estimates of about US$500M are not used. Midjourney publishes no financial statements. Share = the central US$ figure ÷ the part total of US$3,758M. That is the screening total (the register candidates' central figures of US$3,608M plus an outside-register allowance of US$150M for other developers not in the register), held fixed so every company in this sublayer is measured against one total; the card's own estimate for this part is US$2,140M-3,590M-5,660M (low-central-high). Low and high shares hold the denominator at its central."},"avatar-video":{"participates":false,"evidence":{"url":"https://winbuzzer.com/2026/03/31/midjourney-revenue-above-200-million-hardware-push-xcxwbn/","title":"Midjourney Revenue Tops $200M as It Eyes Hardware","publisher":"WinBuzzer","publishedDate":"2026-03-31","accessedDate":"2026-10-22","tier":"T3","locator":"WinBuzzer describes Midjourney as an image (and video) generator sold by subscription; no speech, music, avatar or robot models."}},"robotics-science":{"participates":false,"evidence":{"url":"https://winbuzzer.com/2026/03/31/midjourney-revenue-above-200-million-hardware-push-xcxwbn/","title":"Midjourney Revenue Tops $200M as It Eyes Hardware","publisher":"WinBuzzer","publishedDate":"2026-03-31","accessedDate":"2026-10-22","tier":"T3","locator":"WinBuzzer describes Midjourney as an image (and video) generator sold by subscription; no speech, music, avatar or robot models."}}},"baselineRevenueUsd":{"low":300000000,"central":300000000,"high":300000000,"unit":"USD","basis":"Revenue, calendar year 2024, US dollars (press report)","period":"CY2024 revenue","periodEnd":"2024-12-31","source":{"url":"https://winbuzzer.com/2026/03/31/midjourney-revenue-above-200-million-hardware-push-xcxwbn/","title":"Midjourney Revenue Tops $200M as It Eyes Hardware","publisher":"WinBuzzer","publishedDate":"2026-03-31","tier":"T3","locator":"Article text: \"Per Time's profile of Holz, revenue pushed to $300 million by 2024.\" WinBuzzer relays TIME's profile of David Holz; the figure is TIME's reporting, not WinBuzzer's or the company's","accessedDate":"2026-10-23"},"note":"T3: the US$300M is TIME's reporting, relayed by WinBuzzer (31 March 2026), which names TIME's profile of Holz as its source and does not link it. TIME's own profile (27 August 2025, https://time.com/collections/time100-ai-2025/7305883/david-holz/) words it 'reportedly' and names no source for it; it is not the cited source because the check pass could not open that page (HTTP 406). 'By 2024' does not say whether this is calendar-2024 revenue or a run-rate reached during 2024; it is treated as CY2024. Midjourney publishes no financial statements. More than 12 months old, so the engine widens it; the CEO's own 2026 statement says revenue has kept growing since 2023 without a figure."},"screening":{"image-video":{"levelLow":1,"levelHigh":3}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"d259fe9ecfedb1ae65953944314d09d2d6705073ebe995e54822fe381f4508dd"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"d259fe9ecfedb1ae65953944314d09d2d6705073ebe995e54822fe381f4508dd"}}},"adobe-firefly":{"schema":"5lc-index-inputs/1","key":"models-specialist-models","entityId":"adobe-firefly","name":"Adobe Firefly","kind":"unit","parentId":"adobe","commercialActivityVerified":true,"shares":{"speech-voice":{"low":0,"central":0.1597,"high":1.0644,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Estimate for October 2026","periodEnd":"2026-10-22","source":{"url":"https://www.fool.com/earnings/call-transcripts/2026/09/11/adobe-adbe-q3-2026-earnings-call-transcript/","title":"Adobe (ADBE) Q3 2026 Earnings Call Transcript","publisher":"The Motley Fool (transcript of Adobe's Q3 fiscal 2026 earnings call, 10 September 2026)","publishedDate":"2026-09-11","tier":"T5","locator":"Steven Day: \"Firefly ending ARR across Firefly apps and Firefly credit packs grew 40% quarter over quarter\"; Anil Chakravarthy: \"we added native generation of music, speech, and sound effects\"","accessedDate":"2026-10-22"},"note":"5LayerCake estimate (T5). On its Q3 FY26 call (10 September 2026) Adobe said it had added native generation of music, speech and sound effects to Firefly, and on the Q2 call it cited strong credit consumption driven by video and audio. It gives no audio revenue. Allowance for undisclosed revenue, not a measured figure: central US$3M, high US$20M. Share = the central US$ figure ÷ the part total of US$1,879M. That is the screening total (the register candidates' central figures of US$1,779M plus an outside-register allowance of US$100M for other developers not in the register), held fixed so every company in this sublayer is measured against one total; the card's own estimate for this part is US$1,050M-1,440M-2,280M (low-central-high). Low and high shares hold the denominator at its central."},"music":{"low":0,"central":0.7813,"high":5.2083,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Estimate for October 2026","periodEnd":"2026-10-22","source":{"url":"https://www.fool.com/earnings/call-transcripts/2026/09/11/adobe-adbe-q3-2026-earnings-call-transcript/","title":"Adobe (ADBE) Q3 2026 Earnings Call Transcript","publisher":"The Motley Fool (transcript of Adobe's Q3 fiscal 2026 earnings call, 10 September 2026)","publishedDate":"2026-09-11","tier":"T5","locator":"Steven Day: \"Firefly ending ARR across Firefly apps and Firefly credit packs grew 40% quarter over quarter\"; Anil Chakravarthy: \"we added native generation of music, speech, and sound effects\"","accessedDate":"2026-10-22"},"note":"5LayerCake estimate (T5). Same Q3 FY26 statement as for speech: native generation of music, speech and sound effects was added to Firefly; no audio revenue is given. Allowance for undisclosed revenue, not a measured figure: central US$3M, high US$20M. Share = the central US$ figure ÷ the part total of US$384M. That is the screening total (the register candidates' central figures of US$364M plus an outside-register allowance of US$20M for other developers not in the register), held fixed so every company in this sublayer is measured against one total; the card's own estimate for this part is US$315M-336M-500M (low-central-high). Low and high shares hold the denominator at its central."},"image-video":{"low":4.3906,"central":7.1847,"high":9.5796,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Firefly ARR, fiscal Q2 2026, scaled for Q3 growth","periodEnd":"2026-08-28","source":{"url":"https://www.fool.com/earnings/call-transcripts/2026/06/11/adobe-adbe-q2-2026-earnings-transcript/","title":"Adobe (ADBE) Q2 2026 Earnings Transcript","publisher":"The Motley Fool (transcript of Adobe's Q2 fiscal 2026 earnings call, 11 June 2026)","publishedDate":"2026-06-11","tier":"T5","locator":"David Wadhwani: \"Firefly ending ARR across Firefly app Firefly credit packs, and Firefly enterprise is approaching $300 million exiting Q2.\"","accessedDate":"2026-10-22"},"note":"5LayerCake estimate (T5) from company statements (T2). Adobe's own statements on the earnings calls: Q1 FY26 (quarter ended 27 February 2026) Firefly ending ARR across the Firefly app, credit packs and Firefly Enterprise exceeded US$250M (Q1 call, https://www.fool.com/earnings/call-transcripts/2026/03/12/adobe-adbe-q1-2026-earnings-call-transcript/); Q2 FY26 (29 May 2026) it was approaching US$300M (Q2 call, https://www.fool.com/earnings/call-transcripts/2026/06/11/adobe-adbe-q2-2026-earnings-transcript/); Q3 FY26 (28 August 2026) the app and credit-pack part grew 40% on the quarter with no dollar figure (Q3 call, https://www.fool.com/earnings/call-transcripts/2026/09/11/adobe-adbe-q3-2026-earnings-call-transcript/). Firefly also resells partner models (the Q2 call names Kling 3.0 and Kling 3.0 Omni), so Adobe's own models are taken as 55-90% of the ARR (central 75%) and growth in the Q3 quarter as 0-35% on the total: low 300 × 0.55 = 165; central 300 × 1.2 × 0.75 = 270; high 400 × 0.9 = 360 (US$M). Creative Cloud features count in Applications. Unit figure, never Adobe's total. Share = the central US$ figure ÷ the part total of US$3,758M. That is the screening total (the register candidates' central figures of US$3,608M plus an outside-register allowance of US$150M for other developers not in the register), held fixed so every company in this sublayer is measured against one total; the card's own estimate for this part is US$2,140M-3,590M-5,660M (low-central-high). Low and high shares hold the denominator at its central."},"avatar-video":{"participates":false,"evidence":{"url":"https://www.fool.com/earnings/call-transcripts/2026/09/11/adobe-adbe-q3-2026-earnings-call-transcript/","title":"Adobe (ADBE) Q3 2026 Earnings Call Transcript","publisher":"The Motley Fool (transcript of Adobe's Q3 fiscal 2026 earnings call, 10 September 2026)","publishedDate":"2026-09-11","tier":"T2","locator":"The Q3 FY26 call describes Firefly as image, video and audio generation plus Firefly Enterprise (services, Foundry, Brand Intelligence); it names no avatar-video or robot-model product","accessedDate":"2026-10-22"}},"robotics-science":{"participates":false,"evidence":{"url":"https://www.fool.com/earnings/call-transcripts/2026/09/11/adobe-adbe-q3-2026-earnings-call-transcript/","title":"Adobe (ADBE) Q3 2026 Earnings Call Transcript","publisher":"The Motley Fool (transcript of Adobe's Q3 fiscal 2026 earnings call, 10 September 2026)","publishedDate":"2026-09-11","tier":"T2","locator":"The Q3 FY26 call describes Firefly as image, video and audio generation plus Firefly Enterprise (services, Foundry, Brand Intelligence); it names no avatar-video or robot-model product","accessedDate":"2026-10-22"}}},"baselineRevenueUsd":{"low":295000000,"central":295000000,"high":295000000,"unit":"USD","basis":"Firefly ending ARR (Firefly app, credit packs and Firefly Enterprise), US dollars (company statement)","period":"Firefly ending ARR, fiscal Q2 2026","periodEnd":"2026-05-29","source":{"url":"https://www.fool.com/earnings/call-transcripts/2026/06/11/adobe-adbe-q2-2026-earnings-transcript/","title":"Adobe (ADBE) Q2 2026 Earnings Transcript","publisher":"The Motley Fool (transcript of Adobe's Q2 fiscal 2026 earnings call, 11 June 2026)","publishedDate":"2026-06-11","tier":"T2","locator":"David Wadhwani: \"Firefly ending ARR across Firefly app Firefly credit packs, and Firefly enterprise is approaching $300 million exiting Q2.\"","accessedDate":"2026-10-22"},"note":"The company says Firefly ending ARR was 'approaching $300 million' exiting Q2 FY26; US$295M is taken as the figure. It includes partner-model usage billed through Firefly. The Firefly unit only, never Adobe's total ARR."},"screening":{"speech-voice":{"levelLow":1,"levelHigh":2},"music":{"levelLow":1,"levelHigh":2},"image-video":{"levelLow":1,"levelHigh":3}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"2da25aa03763e782b8a08d1b3641bce934d385ab9e89a501f3e7b90b8f7d743f"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"2da25aa03763e782b8a08d1b3641bce934d385ab9e89a501f3e7b90b8f7d743f"}}},"openai":{"schema":"5lc-index-inputs/1","key":"models-specialist-models","entityId":"openai","name":"OpenAI (image, video and audio models)","kind":"private","commercialActivityVerified":true,"shares":{"speech-voice":{"low":1.0646,"central":4.2583,"high":13.3071,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Estimate for October 2026","periodEnd":"2026-10-22","source":{"url":"https://thenextweb.com/news/openai-gpt-realtime-2-voice-models","title":"OpenAI releases GPT-Realtime-2, GPT-Realtime-Translate and GPT-Realtime-Whisper voice models in its API","publisher":"The Next Web","publishedDate":"2026-05-08","locator":"Three voice models in the API: GPT-Realtime-2 ($32 / $64 per million audio tokens), Realtime-Translate ($0.034 a minute), Realtime-Whisper ($0.017 a minute); named early users Zillow, Glean, Intercom, Priceline, Vimeo and Deutsche Telekom; no usage or revenue figures","tier":"T5","accessedDate":"2026-10-22"},"note":"5LayerCake estimate (T5). OpenAI sells GPT-Realtime-2 ($32 / $64 per million audio tokens), Realtime-Translate ($0.034 a minute) and Realtime-Whisper ($0.017 a minute) with named customers (Zillow, Deutsche Telekom, Priceline, Vimeo) but no usage or revenue figures. Share of the card's large-company speech allowance: 80 of 600 (US$M). Real-time voice inside the general models counts in Frontier labs. Share = the central US$ figure ÷ the part total of US$1,879M, which is the register candidates' central figures (US$1,779M) plus an outside-register allowance of US$100M for other speech developers not in the register (for example Gladia, Soniox, Fish Audio, Murf, WellSaid, Rev); the card's own estimate for this part is US$1,050M-1,440M-2,280M (low-central-high). Low and high shares hold the denominator at its central."},"music":{"low":0,"central":0.2602,"high":2.6021,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Estimate for October 2026","periodEnd":"2026-10-22","source":{"url":"https://thenextweb.com/news/openai-gpt-realtime-2-voice-models","title":"OpenAI releases GPT-Realtime-2, GPT-Realtime-Translate and GPT-Realtime-Whisper voice models in its API","publisher":"The Next Web","publishedDate":"2026-05-08","locator":"Three voice models in the API: GPT-Realtime-2 ($32 / $64 per million audio tokens), Realtime-Translate ($0.034 a minute), Realtime-Whisper ($0.017 a minute); named early users Zillow, Glean, Intercom, Priceline, Vimeo and Deutsche Telekom; no usage or revenue figures","tier":"T5","accessedDate":"2026-10-22"},"note":"5LayerCake estimate (T5). No OpenAI music product or revenue was found in the cited article. Allowance for undisclosed revenue, not a measured figure. Share = the central US$ figure ÷ the part total of US$384M, which is the register candidates' central figures (US$364M) plus an outside-register allowance of US$20M for other music-model developers not in the register (for example Producer.ai, AIVA, Soundraw); the card's own estimate for this part is US$315M-336M-500M (low-central-high). Low and high shares hold the denominator at its central."},"image-video":{"low":1.5965,"central":6.6519,"high":15.9647,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Estimate for October 2026","periodEnd":"2026-09-17","source":{"url":"https://vercel.com/blog/ai-gateway-production-index-september-2026","title":"Open-weight models take 56% of token volume, Astra doubles Fable 5.1 spend (AI Gateway Production Index, September 2026)","publisher":"Vercel","publishedDate":"2026-09-17","accessedDate":"2026-10-22","tier":"T5","locator":"August 2026 data: Nano Banana 50% of image spend vs GPT Image 44%; Seedance first in videos and video spend; Veo 20% of video spend; Grok Imagine 19% of videos"},"note":"5LayerCake estimate (T5). OpenAI does not disclose API revenue for images. Same split of the card's US$800M allowance as for Google: GPT Image was 44-46% of gateway image traffic in August 2026 (Nano Banana 50% of image spend). OpenAI about 31% = US$250M, range 60-600M. ChatGPT image features count in Applications; any Sora API revenue is not separately visible. Share = the central US$ figure ÷ the part total of US$3,758M, which is the register candidates' central figures (US$3,608M) plus an outside-register allowance of US$150M for other image and video developers not in the register (for example Higgsfield, Freepik, Tripo, Genmo); the card's own estimate for this part is US$2,140M-3,590M-5,660M (low-central-high). Low and high shares hold the denominator at its central."},"avatar-video":{"low":0,"central":0.2011,"high":2.0109,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Estimate for October 2026","periodEnd":"2026-10-22","source":{"url":"https://vercel.com/blog/ai-gateway-production-index-september-2026","title":"Open-weight models take 56% of token volume, Astra doubles Fable 5.1 spend (AI Gateway Production Index, September 2026)","publisher":"Vercel","publishedDate":"2026-09-17","accessedDate":"2026-10-22","tier":"T5","locator":"August 2026 data: Nano Banana 50% of image spend vs GPT Image 44%; Seedance first in videos and video spend; Veo 20% of video spend; Grok Imagine 19% of videos"},"note":"5LayerCake estimate (T5). No OpenAI avatar-video product or revenue was found in the cited page. Allowance for undisclosed revenue, not a measured figure. Share = the central US$ figure ÷ the part total of US$497M, which is the register candidates' central figures (US$467M) plus an outside-register allowance of US$30M for other avatar-video developers not in the register (for example Akool, Argil, Creatify); the card's own estimate for this part is US$400M-450M-550M (low-central-high). Low and high shares hold the denominator at its central."},"robotics-science":{"low":0,"central":0.2139,"high":2.1395,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Estimate for October 2026","periodEnd":"2026-10-22","source":{"url":"https://thenextweb.com/news/openai-gpt-realtime-2-voice-models","title":"OpenAI releases GPT-Realtime-2, GPT-Realtime-Translate and GPT-Realtime-Whisper voice models in its API","publisher":"The Next Web","publishedDate":"2026-05-08","locator":"Three voice models in the API: GPT-Realtime-2 ($32 / $64 per million audio tokens), Realtime-Translate ($0.034 a minute), Realtime-Whisper ($0.017 a minute); named early users Zillow, Glean, Intercom, Priceline, Vimeo and Deutsche Telekom; no usage or revenue figures","tier":"T5","accessedDate":"2026-10-22"},"note":"5LayerCake estimate (T5). No OpenAI robot-model product or revenue was found in the cited article. Allowance for undisclosed revenue, not a measured figure. Share = the central US$ figure ÷ the part total of US$234M, which is the register candidates' central figures (US$219M) plus an outside-register allowance of US$15M for other robot-model and science-model developers not in the register (for example Latent Labs, Periodic Labs, Absci); the card's own estimate for this part is US$140M-200M-350M (low-central-high). Low and high shares hold the denominator at its central."}},"screening":{"speech-voice":{"levelLow":1,"levelHigh":2},"music":{"levelLow":1,"levelHigh":2},"image-video":{"levelLow":1,"levelHigh":3},"avatar-video":{"levelLow":1,"levelHigh":2},"robotics-science":{"levelLow":1,"levelHigh":2}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"a1234debe90be02859939e8077114dd6a8066c8e2fd3e295de4e3856e06afa76"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"a1234debe90be02859939e8077114dd6a8066c8e2fd3e295de4e3856e06afa76"}}},"heygen":{"schema":"5lc-index-inputs/1","key":"models-specialist-models","entityId":"heygen","name":"HeyGen","kind":"private","commercialActivityVerified":true,"shares":{"speech-voice":{"participates":false,"evidence":{"url":"https://www.heygen.com/blog/heygen-surpasses-200m-arr","title":"HeyGen Doubles to $200M ARR in Eight Months on the Rise of Identity-First AI Video","publisher":"HeyGen","publishedDate":"2026-06-25","accessedDate":"2026-10-22","tier":"T2","locator":"HeyGen's release describes avatar and AI video creation for business; no speech-model, music, image or robot products."}},"music":{"participates":false,"evidence":{"url":"https://www.heygen.com/blog/heygen-surpasses-200m-arr","title":"HeyGen Doubles to $200M ARR in Eight Months on the Rise of Identity-First AI Video","publisher":"HeyGen","publishedDate":"2026-06-25","accessedDate":"2026-10-22","tier":"T2","locator":"HeyGen's release describes avatar and AI video creation for business; no speech-model, music, image or robot products."}},"image-video":{"participates":false,"evidence":{"url":"https://www.heygen.com/blog/heygen-surpasses-200m-arr","title":"HeyGen Doubles to $200M ARR in Eight Months on the Rise of Identity-First AI Video","publisher":"HeyGen","publishedDate":"2026-06-25","accessedDate":"2026-10-22","tier":"T2","locator":"HeyGen's release describes avatar and AI video creation for business; no speech-model, music, image or robot products."}},"avatar-video":{"low":40.2414,"central":40.2414,"high":40.2414,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"ARR, June 2026 (surpassed US$200M on 25 June 2026)","periodEnd":"2026-06-30","source":{"url":"https://www.heygen.com/blog/heygen-surpasses-200m-arr","title":"HeyGen Doubles to $200M ARR in Eight Months on the Rise of Identity-First AI Video","publisher":"HeyGen","publishedDate":"2026-06-25","accessedDate":"2026-10-22","tier":"T2","locator":"Release: HeyGen surpassed $200 million in annual recurring revenue, doubling in eight months (dateline 25 June 2026)"},"note":"The company says it surpassed US$200M of ARR, doubling in eight months (release dated 25 June 2026); 30M+ users, 85% of the Fortune 100. 'Surpassed' means at least this, and the figure is taken as at the June 2026 month end. Sacra puts June 2026 at US$205M (T4, same series as the prior below). HeyGen is counted here, not in Vertical AI. Share = the central US$ figure ÷ the part total of US$497M. That is the screening total (the register candidates' central figures of US$467M plus an outside-register allowance of US$30M for other developers not in the register), held fixed so every company in this sublayer is measured against one total; the card's own estimate for this part is US$400M-450M-550M (low-central-high). Low and high shares hold the denominator at its central."},"robotics-science":{"participates":false,"evidence":{"url":"https://www.heygen.com/blog/heygen-surpasses-200m-arr","title":"HeyGen Doubles to $200M ARR in Eight Months on the Rise of Identity-First AI Video","publisher":"HeyGen","publishedDate":"2026-06-25","accessedDate":"2026-10-22","tier":"T2","locator":"HeyGen's release describes avatar and AI video creation for business; no speech-model, music, image or robot products."}}},"priorShares":{"avatar-video":{"low":34.7887,"central":34.7887,"high":34.7887,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"ARR, September 2025 (Sacra estimate)","periodEnd":"2025-09-30","source":{"url":"https://sacra.com/c/heygen/","title":"HeyGen (Sacra company profile)","publisher":"Sacra","publishedDate":"undated","tier":"T4","accessedDate":"2026-10-22","locator":"Competition section: 'HeyGen at $95M ARR in September 2025'; the same page estimates US$205M in June 2026, US$123M at end-2025 and US$57.5M at end-2024"},"note":"Sacra's estimate (T4); the page states no method. The company's own release gives no earlier dated figure (it says only 'doubling in eight months', i.e. about US$100M in October 2025, which is under 9 months before the current figure and so cannot be used as the prior). Sacra's September 2025 value is on the same series as its June 2026 value. Prior share = the prior US$ figure ÷ the prior part total of US$273M, which is the current part total of US$497M ÷ 1.82, the growth of this part over the year: pooled from the two avatar companies with a sourced pair (HeyGen US$205M ÷ US$95M, Synthesia US$150M ÷ US$100M), US$355M ÷ US$195M = 1.82. Companies without a sourced prior are assumed to have grown at that rate. Low and high hold the denominator at its central."}},"activityGrowth":{"current":{"low":205000000,"central":205000000,"high":205000000,"unit":"USD","basis":"Annual recurring revenue (ARR) of HeyGen, US dollars (Sacra estimate)","period":"ARR, June 2026 (Sacra estimate)","periodEnd":"2026-06-30","source":{"url":"https://sacra.com/c/heygen/","title":"HeyGen (Sacra company profile)","publisher":"Sacra","publishedDate":"undated","tier":"T4","accessedDate":"2026-10-22","locator":"'Sacra estimates that HeyGen hit $205M in annual recurring revenue (ARR) in June 2026, up from $123M at the end of 2025'"},"note":"Sacra estimate (T4), within 3% of the company's own 'surpassed US$200M' (25 June 2026, T2)."},"prior":{"low":95000000,"central":95000000,"high":95000000,"unit":"USD","basis":"Annual recurring revenue (ARR) of HeyGen, US dollars (Sacra estimate)","period":"ARR, September 2025 (Sacra estimate)","periodEnd":"2025-09-30","source":{"url":"https://sacra.com/c/heygen/","title":"HeyGen (Sacra company profile)","publisher":"Sacra","publishedDate":"undated","tier":"T4","accessedDate":"2026-10-22","locator":"Competition section: 'HeyGen at $95M ARR in September 2025'"},"note":"Same Sacra series. Growth 205 ÷ 95 = 2.2 times in nine months."}},"baselineRevenueUsd":{"low":200000000,"central":200000000,"high":200000000,"unit":"USD","basis":"Annual recurring revenue (ARR) of HeyGen, US dollars (company statement)","period":"ARR, June 2026","periodEnd":"2026-06-30","source":{"url":"https://www.heygen.com/blog/heygen-surpasses-200m-arr","title":"HeyGen Doubles to $200M ARR in Eight Months on the Rise of Identity-First AI Video","publisher":"HeyGen","publishedDate":"2026-06-25","accessedDate":"2026-10-22","tier":"T2","locator":"Release: HeyGen surpassed $200 million in annual recurring revenue, doubling in eight months (dateline 25 June 2026)"},"note":"The company says it surpassed US$200M of ARR (25 June 2026). ARR is a run-rate."},"screening":{"avatar-video":{"levelLow":1,"levelHigh":3}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"c6ac31596140a87ac6e4b007cb5ada782b5bb21df6a6901ec4809b6ded422035"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"c6ac31596140a87ac6e4b007cb5ada782b5bb21df6a6901ec4809b6ded422035"}}},"skild-ai":{"schema":"5lc-index-inputs/1","key":"models-specialist-models","entityId":"skild-ai","name":"Skild AI","kind":"private","commercialActivityVerified":true,"shares":{"speech-voice":{"participates":false,"evidence":{"url":"https://www.skild.ai/blogs/skild-crosses-100m-arr","title":"The Hidden Pillar of Robotics","publisher":"Skild AI","publishedDate":"2026-09-10","accessedDate":"2026-10-22","tier":"T2","locator":"Skild's post describes robot manipulation and mobility models; no speech, music, image, video or avatar products."}},"music":{"participates":false,"evidence":{"url":"https://www.skild.ai/blogs/skild-crosses-100m-arr","title":"The Hidden Pillar of Robotics","publisher":"Skild AI","publishedDate":"2026-09-10","accessedDate":"2026-10-22","tier":"T2","locator":"Skild's post describes robot manipulation and mobility models; no speech, music, image, video or avatar products."}},"image-video":{"participates":false,"evidence":{"url":"https://www.skild.ai/blogs/skild-crosses-100m-arr","title":"The Hidden Pillar of Robotics","publisher":"Skild AI","publishedDate":"2026-09-10","accessedDate":"2026-10-22","tier":"T2","locator":"Skild's post describes robot manipulation and mobility models; no speech, music, image, video or avatar products."}},"avatar-video":{"participates":false,"evidence":{"url":"https://www.skild.ai/blogs/skild-crosses-100m-arr","title":"The Hidden Pillar of Robotics","publisher":"Skild AI","publishedDate":"2026-09-10","accessedDate":"2026-10-22","tier":"T2","locator":"Skild's post describes robot manipulation and mobility models; no speech, music, image, video or avatar products."}},"robotics-science":{"low":42.7899,"central":42.7899,"high":42.7899,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"ARR, September 2026","periodEnd":"2026-09-10","source":{"url":"https://www.skild.ai/blogs/skild-crosses-100m-arr","title":"The Hidden Pillar of Robotics","publisher":"Skild AI","publishedDate":"2026-09-10","accessedDate":"2026-10-22","tier":"T2","locator":"Crossed $100M ARR; $50M recognised in 10 months of deployments; 60+ paying customers; about 90% manipulation, 10% mobility"},"note":"The company says it has crossed US$100M ARR, US$50M recognised in about 10 months of deployments, 60+ paying customers; about 90% manipulation and 10% mobility (post of 10 Sept 2026). Skild's revenue includes robot-model licences and deployments; the AMR portion (4% of revenue) is from the Zebra/Fetch acquisition. Share = the central US$ figure ÷ the part total of US$234M, which is the register candidates' central figures (US$219M) plus an outside-register allowance of US$15M for other robot-model and science-model developers not in the register (for example Latent Labs, Periodic Labs, Absci); the card's own estimate for this part is US$140M-200M-350M (low-central-high). Low and high shares hold the denominator at its central."}},"priorShares":{"robotics-science":{"low":0,"central":0,"high":0,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Before first commercial deployment (September 2025)","periodEnd":"2025-09-10","source":{"url":"https://www.skild.ai/blogs/skild-crosses-100m-arr","title":"The Hidden Pillar of Robotics","publisher":"Skild AI","publishedDate":"2026-09-10","accessedDate":"2026-10-22","tier":"T2","locator":"Post of 10 September 2026: $100M ARR 'ten months after our first commercial deployment'; '$50M has already been recognized since starting deployments 10 months ago'"},"note":"The company says it reached US$100M of ARR ten months after its first commercial deployment and has recognised US$50M since starting deployments 10 months ago, so its commercial deployments began around November 2025 and it had no deployment revenue in September 2025. Zero is therefore an observation for that date, not a missing value."}},"activityGrowth":{"current":{"low":100000000,"central":100000000,"high":100000000,"unit":"USD","basis":"Annual recurring revenue (ARR) of Skild AI, US dollars (company statement)","period":"ARR, September 2026","periodEnd":"2026-09-10","source":{"url":"https://www.skild.ai/blogs/skild-crosses-100m-arr","title":"The Hidden Pillar of Robotics","publisher":"Skild AI","publishedDate":"2026-09-10","accessedDate":"2026-10-22","tier":"T2","locator":"Crossed $100M ARR; $50M recognised in 10 months of deployments; 60+ paying customers; about 90% manipulation, 10% mobility"},"note":"The company says: 'We have crossed $100M in annual recurring revenue.'"},"prior":{"low":0,"central":0,"high":0,"unit":"USD","basis":"Annual recurring revenue (ARR) of Skild AI, US dollars (company statement)","period":"Before first commercial deployment (September 2025)","periodEnd":"2025-09-10","source":{"url":"https://www.skild.ai/blogs/skild-crosses-100m-arr","title":"The Hidden Pillar of Robotics","publisher":"Skild AI","publishedDate":"2026-09-10","accessedDate":"2026-10-22","tier":"T2","locator":"Post of 10 September 2026: $100M ARR 'ten months after our first commercial deployment'"},"note":"No commercial deployment, so no ARR, ten months before the post (see the prior share). The engine counts growth from zero as +300%."}},"baselineRevenueUsd":{"low":100000000,"central":100000000,"high":100000000,"unit":"USD","basis":"Annual recurring revenue (ARR) of Skild AI, US dollars (company statement)","period":"ARR, September 2026","periodEnd":"2026-09-10","source":{"url":"https://www.skild.ai/blogs/skild-crosses-100m-arr","title":"The Hidden Pillar of Robotics","publisher":"Skild AI","publishedDate":"2026-09-10","accessedDate":"2026-10-22","tier":"T2","locator":"Crossed $100M ARR; $50M recognised in 10 months of deployments; 60+ paying customers; about 90% manipulation, 10% mobility"},"note":"The company says it has crossed US$100M of ARR; US$50M recognised so far. ARR is a run-rate."},"screening":{"robotics-science":{"levelLow":2,"levelHigh":4}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"c7ddb59d09aad4ef06e347b7c1d3ab6830d343dca55646ad9d2ba7e7fff0e3aa"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"c7ddb59d09aad4ef06e347b7c1d3ab6830d343dca55646ad9d2ba7e7fff0e3aa"}}},"synthesia":{"schema":"5lc-index-inputs/1","key":"models-specialist-models","entityId":"synthesia","name":"Synthesia","kind":"private","commercialActivityVerified":true,"shares":{"speech-voice":{"participates":false,"evidence":{"url":"https://www.synthesia.io/post/series-e-200-million-4-billion-valuation-future-work","title":"Synthesia raises $200 million Series E at $4 billion valuation to change how companies train and upskill their workforce","publisher":"Synthesia","publishedDate":"2026-01-26","accessedDate":"2026-10-22","tier":"T2","locator":"Synthesia's Series E post describes AI avatar video for workforce training; no speech-model, music, image or robot products."}},"music":{"participates":false,"evidence":{"url":"https://www.synthesia.io/post/series-e-200-million-4-billion-valuation-future-work","title":"Synthesia raises $200 million Series E at $4 billion valuation to change how companies train and upskill their workforce","publisher":"Synthesia","publishedDate":"2026-01-26","accessedDate":"2026-10-22","tier":"T2","locator":"Synthesia's Series E post describes AI avatar video for workforce training; no speech-model, music, image or robot products."}},"image-video":{"participates":false,"evidence":{"url":"https://www.synthesia.io/post/series-e-200-million-4-billion-valuation-future-work","title":"Synthesia raises $200 million Series E at $4 billion valuation to change how companies train and upskill their workforce","publisher":"Synthesia","publishedDate":"2026-01-26","accessedDate":"2026-10-22","tier":"T2","locator":"Synthesia's Series E post describes AI avatar video for workforce training; no speech-model, music, image or robot products."}},"avatar-video":{"low":30.1629,"central":30.1629,"high":30.1629,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"ARR, January 2026","periodEnd":"2026-01-26","source":{"url":"https://seedcamp.com/views/synthesia-raises-200m-series-e-at-4b-valuation-to-continue-the-next-era-of-video-creation/","title":"Synthesia raises $200M Series E at $4B valuation to continue the next era of video creation","publisher":"Seedcamp (investor)","publishedDate":"2026-01-26","accessedDate":"2026-10-22","tier":"T2","locator":"Investor post: Synthesia has $150 million in annual recurring revenue, over 1 million users and 60,000 businesses"},"note":"Stated as company ARR by its investor Seedcamp (26 Jan 2026): US$150M, with 1M+ users and 60,000 businesses. Synthesia's own Series E page (same day) gives no ARR figure. Counted here, not in Vertical AI. Share = the central US$ figure ÷ the part total of US$497M, which is the register candidates' central figures (US$467M) plus an outside-register allowance of US$30M for other avatar-video developers not in the register (for example Akool, Argil, Creatify); the card's own estimate for this part is US$400M-450M-550M (low-central-high). Low and high shares hold the denominator at its central."},"robotics-science":{"participates":false,"evidence":{"url":"https://www.synthesia.io/post/series-e-200-million-4-billion-valuation-future-work","title":"Synthesia raises $200 million Series E at $4 billion valuation to change how companies train and upskill their workforce","publisher":"Synthesia","publishedDate":"2026-01-26","accessedDate":"2026-10-22","tier":"T2","locator":"Synthesia's Series E post describes AI avatar video for workforce training; no speech-model, music, image or robot products."}}},"priorShares":{"avatar-video":{"low":36.6197,"central":36.6197,"high":36.6197,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"ARR, April 2025 (surpassed US$100M)","periodEnd":"2025-04-15","source":{"url":"https://www.synthesia.io/post/100-million-revenue-adobe-investment","title":"Synthesia surpasses $100 million in annual recurring revenue and secures strategic investment from Adobe Ventures","publisher":"Synthesia","publishedDate":"2025-04-15","tier":"T2","locator":"\"today announced it has reached over $100 million in annual recurring revenue (ARR)\" (dateline London, 15 April 2025)","accessedDate":"2026-10-22"},"note":"The company says it reached over US$100M of ARR (release of 15 April 2025). 'Over' makes this a floor. The current US$150M is the investor's statement of 26 January 2026, 9 months 11 days later. Prior share = the prior US$ figure ÷ the prior part total of US$273M, which is the current part total of US$497M ÷ 1.82, the growth of this part over the year: pooled from the two avatar companies with a sourced pair (HeyGen US$205M ÷ US$95M, Synthesia US$150M ÷ US$100M), US$355M ÷ US$195M = 1.82. Companies without a sourced prior are assumed to have grown at that rate. Low and high hold the denominator at its central."}},"activityGrowth":{"current":{"low":150000000,"central":150000000,"high":150000000,"unit":"USD","basis":"Annual recurring revenue (ARR) of Synthesia, US dollars (company statements)","period":"ARR, January 2026","periodEnd":"2026-01-26","source":{"url":"https://seedcamp.com/views/synthesia-raises-200m-series-e-at-4b-valuation-to-continue-the-next-era-of-video-creation/","title":"Synthesia raises $200M Series E at $4B valuation to continue the next era of video creation","publisher":"Seedcamp (investor)","publishedDate":"2026-01-26","accessedDate":"2026-10-22","tier":"T2","locator":"Investor post: Synthesia has $150 million in annual recurring revenue, over 1 million users and 60,000 businesses"},"note":"Stated as company ARR by its investor Seedcamp (26 January 2026): US$150M. Synthesia's own Series E post of the same day gives no ARR figure."},"prior":{"low":100000000,"central":100000000,"high":100000000,"unit":"USD","basis":"Annual recurring revenue (ARR) of Synthesia, US dollars (company statements)","period":"ARR, April 2025","periodEnd":"2025-04-15","source":{"url":"https://www.synthesia.io/post/100-million-revenue-adobe-investment","title":"Synthesia surpasses $100 million in annual recurring revenue and secures strategic investment from Adobe Ventures","publisher":"Synthesia","publishedDate":"2025-04-15","tier":"T2","locator":"\"today announced it has reached over $100 million in annual recurring revenue (ARR)\" (dateline London, 15 April 2025)","accessedDate":"2026-10-22"},"note":"Growth 150 ÷ 100 = 1.5 times in 9 months 11 days. 'Over US$100M' is a floor, so the ratio is a ceiling."}},"baselineRevenueUsd":{"low":150000000,"central":150000000,"high":150000000,"unit":"USD","basis":"Annual recurring revenue (ARR) of Synthesia, US dollars (company statements)","period":"ARR, January 2026","periodEnd":"2026-01-26","source":{"url":"https://seedcamp.com/views/synthesia-raises-200m-series-e-at-4b-valuation-to-continue-the-next-era-of-video-creation/","title":"Synthesia raises $200M Series E at $4B valuation to continue the next era of video creation","publisher":"Seedcamp (investor)","publishedDate":"2026-01-26","accessedDate":"2026-10-22","tier":"T2","locator":"Investor post: Synthesia has $150 million in annual recurring revenue, over 1 million users and 60,000 businesses"},"note":"Stated by its investor Seedcamp as company ARR; an ARR is a run-rate."},"screening":{"avatar-video":{"levelLow":1,"levelHigh":3}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"08c4dd22c422ee162a69a9fd02849bddeb0b3d47eb08d468cd9bde67f53d0f1c"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"08c4dd22c422ee162a69a9fd02849bddeb0b3d47eb08d468cd9bde67f53d0f1c"}}},"runway":{"schema":"5lc-index-inputs/1","key":"models-specialist-models","entityId":"runway","name":"Runway","kind":"private","commercialActivityVerified":true,"shares":{"speech-voice":{"participates":false,"evidence":{"url":"https://www.pymnts.com/?p=4167359","title":"Runway AI Hits $200 Million Revenue and Expands Robotics Capabilities","publisher":"PYMNTS","publishedDate":"2026-09-08","accessedDate":"2026-10-22","tier":"T3","locator":"PYMNTS describes Runway as a video and world-model company; its Runway Robotics offering is simulation for policy inference, not a robot-control model, so no revenue is counted here."}},"music":{"participates":false,"evidence":{"url":"https://www.pymnts.com/?p=4167359","title":"Runway AI Hits $200 Million Revenue and Expands Robotics Capabilities","publisher":"PYMNTS","publishedDate":"2026-09-08","accessedDate":"2026-10-22","tier":"T3","locator":"PYMNTS describes Runway as a video and world-model company; its Runway Robotics offering is simulation for policy inference, not a robot-control model, so no revenue is counted here."}},"image-video":{"low":5.3216,"central":5.3216,"high":5.3216,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"ARR, September 2026","periodEnd":"2026-09-08","source":{"url":"https://www.pymnts.com/?p=4167359","title":"Runway AI Hits $200 Million Revenue and Expands Robotics Capabilities","publisher":"PYMNTS","publishedDate":"2026-09-08","accessedDate":"2026-10-22","tier":"T2","locator":"Runway reached $200 million ARR, roughly doubling since April (co-CEO Anastasis Germanidis, LinkedIn post)"},"note":"The company says $200M ARR, roughly doubled since April (co-CEO Anastasis Germanidis on LinkedIn, reported by PYMNTS 8 Sept 2026). A year-end target of $350M comes from an anonymous source in secondary coverage and is not used. Share = the central US$ figure ÷ the part total of US$3,758M, which is the register candidates' central figures (US$3,608M) plus an outside-register allowance of US$150M for other image and video developers not in the register (for example Higgsfield, Freepik, Tripo, Genmo); the card's own estimate for this part is US$2,140M-3,590M-5,660M (low-central-high). Low and high shares hold the denominator at its central."},"avatar-video":{"participates":false,"evidence":{"url":"https://www.pymnts.com/?p=4167359","title":"Runway AI Hits $200 Million Revenue and Expands Robotics Capabilities","publisher":"PYMNTS","publishedDate":"2026-09-08","accessedDate":"2026-10-22","tier":"T3","locator":"PYMNTS describes Runway as a video and world-model company; its Runway Robotics offering is simulation for policy inference, not a robot-control model, so no revenue is counted here."}},"robotics-science":{"participates":false,"evidence":{"url":"https://www.pymnts.com/?p=4167359","title":"Runway AI Hits $200 Million Revenue and Expands Robotics Capabilities","publisher":"PYMNTS","publishedDate":"2026-09-08","accessedDate":"2026-10-22","tier":"T3","locator":"PYMNTS describes Runway as a video and world-model company; its Runway Robotics offering is simulation for policy inference, not a robot-control model, so no revenue is counted here."}}},"baselineRevenueUsd":{"low":200000000,"central":200000000,"high":200000000,"unit":"USD","basis":"Annual recurring revenue (ARR) of Runway, US dollars (company statement)","period":"ARR, September 2026","periodEnd":"2026-09-08","source":{"url":"https://www.pymnts.com/?p=4167359","title":"Runway AI Hits $200 Million Revenue and Expands Robotics Capabilities","publisher":"PYMNTS","publishedDate":"2026-09-08","accessedDate":"2026-10-22","tier":"T2","locator":"Runway reached $200 million ARR, roughly doubling since April (co-CEO Anastasis Germanidis, LinkedIn post)"},"note":"The company says ARR has doubled since April to US$200M (co-CEO Anastasis Germanidis, LinkedIn post of 8 September 2026, reported by PYMNTS). Sacra estimates US$96M at the end of 2025, US$100M in April and US$200M in August 2026 (https://sacra.com/c/runway/, T4); its end-2025 value is 8 months 8 days before the current date, outside the 9 to 15 month window, so no growth component is given."},"screening":{"image-video":{"levelLow":1,"levelHigh":3}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"1213b2f145dec6e70d46193ebf01d5d85bc25b3979eb1bbd57a8ee37cabff529"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"1213b2f145dec6e70d46193ebf01d5d85bc25b3979eb1bbd57a8ee37cabff529"}}},"microsoft-ai":{"schema":"5lc-index-inputs/1","key":"models-specialist-models","entityId":"microsoft-ai","name":"Microsoft AI (MAI) and Azure AI Speech","kind":"unit","parentId":"microsoft","commercialActivityVerified":true,"shares":{"speech-voice":{"low":3.1937,"central":9.0488,"high":21.2913,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Estimate for October 2026","periodEnd":"2026-04-02","source":{"url":"https://microsoft.ai/news/today-were-announcing-3-new-world-class-mai-models-available-in-foundry/","title":"Announcing 3 new world class MAI models, available in Foundry","publisher":"Microsoft AI","publishedDate":"2026-04-02","accessedDate":"2026-10-22","tier":"T5","locator":"MAI-Transcribe-1 ($0.36 an hour), MAI-Voice-1 ($22 per million characters) and MAI-Image-2 sold through Microsoft Foundry; WPP named as a first enterprise partner; no revenue given"},"note":"5LayerCake estimate (T5). MAI-Voice-1 ($22 per million characters), MAI-Transcribe-1 ($0.36 an hour) and Azure AI Speech are sold through Foundry; no revenue is disclosed. Share of the card's US$600M large-company speech allowance: 170. Unit figure, not Microsoft's total. Share = the central US$ figure ÷ the part total of US$1,879M, which is the register candidates' central figures (US$1,779M) plus an outside-register allowance of US$100M for other speech developers not in the register (for example Gladia, Soniox, Fish Audio, Murf, WellSaid, Rev); the card's own estimate for this part is US$1,050M-1,440M-2,280M (low-central-high). Low and high shares hold the denominator at its central."},"music":{"low":0,"central":0.1301,"high":1.3011,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Estimate for October 2026","periodEnd":"2026-10-22","source":{"url":"https://microsoft.ai/news/today-were-announcing-3-new-world-class-mai-models-available-in-foundry/","title":"Announcing 3 new world class MAI models, available in Foundry","publisher":"Microsoft AI","publishedDate":"2026-04-02","accessedDate":"2026-10-22","tier":"T5","locator":"MAI-Transcribe-1 ($0.36 an hour), MAI-Voice-1 ($22 per million characters) and MAI-Image-2 sold through Microsoft Foundry; WPP named as a first enterprise partner; no revenue given"},"note":"5LayerCake estimate (T5). The three MAI models launched in April 2026 are speech, voice and image; no music product was found. Allowance for undisclosed revenue, not a measured figure. Share = the central US$ figure ÷ the part total of US$384M, which is the register candidates' central figures (US$364M) plus an outside-register allowance of US$20M for other music-model developers not in the register (for example Producer.ai, AIVA, Soundraw); the card's own estimate for this part is US$315M-336M-500M (low-central-high). Low and high shares hold the denominator at its central."},"image-video":{"low":0.133,"central":0.7982,"high":2.6608,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Estimate for October 2026","periodEnd":"2026-04-02","source":{"url":"https://microsoft.ai/news/today-were-announcing-3-new-world-class-mai-models-available-in-foundry/","title":"Announcing 3 new world class MAI models, available in Foundry","publisher":"Microsoft AI","publishedDate":"2026-04-02","accessedDate":"2026-10-22","tier":"T5","locator":"MAI-Transcribe-1 ($0.36 an hour), MAI-Voice-1 ($22 per million characters) and MAI-Image-2 sold through Microsoft Foundry; WPP named as a first enterprise partner; no revenue given"},"note":"5LayerCake estimate (T5). MAI-Image-2 is sold through Microsoft Foundry ($5 / $33 per million tokens, WPP a named partner); no revenue is disclosed. Share of the card's US$800M allowance for the four large companies: about 4%. Unit figure, not Microsoft's total. Share = the central US$ figure ÷ the part total of US$3,758M, which is the register candidates' central figures (US$3,608M) plus an outside-register allowance of US$150M for other image and video developers not in the register (for example Higgsfield, Freepik, Tripo, Genmo); the card's own estimate for this part is US$2,140M-3,590M-5,660M (low-central-high). Low and high shares hold the denominator at its central."},"avatar-video":{"low":0,"central":0.4022,"high":2.413,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Estimate for October 2026","periodEnd":"2026-10-22","source":{"url":"https://azure.microsoft.com/en-us/pricing/details/cognitive-services/speech-services/","title":"Azure Speech in Foundry Tools pricing","publisher":"Microsoft","publishedDate":"undated","accessedDate":"2026-10-22","tier":"T5","locator":"Price list for speech to text, text to speech (including MAI-Voice) and speech translation; dollar amounts not shown in the copy read"},"note":"5LayerCake estimate (T5). Azure's speech price list includes avatar add-ons per minute; no revenue disclosed. Allowance for undisclosed revenue, not a measured figure. Share = the central US$ figure ÷ the part total of US$497M, which is the register candidates' central figures (US$467M) plus an outside-register allowance of US$30M for other avatar-video developers not in the register (for example Akool, Argil, Creatify); the card's own estimate for this part is US$400M-450M-550M (low-central-high). Low and high shares hold the denominator at its central."},"robotics-science":{"low":0,"central":0.1284,"high":1.2837,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Estimate for October 2026","periodEnd":"2026-10-22","source":{"url":"https://microsoft.ai/news/today-were-announcing-3-new-world-class-mai-models-available-in-foundry/","title":"Announcing 3 new world class MAI models, available in Foundry","publisher":"Microsoft AI","publishedDate":"2026-04-02","accessedDate":"2026-10-22","tier":"T5","locator":"MAI-Transcribe-1 ($0.36 an hour), MAI-Voice-1 ($22 per million characters) and MAI-Image-2 sold through Microsoft Foundry; WPP named as a first enterprise partner; no revenue given"},"note":"5LayerCake estimate (T5). No Microsoft robot-model product or revenue was found in the cited page. Allowance for undisclosed revenue, not a measured figure. Share = the central US$ figure ÷ the part total of US$234M, which is the register candidates' central figures (US$219M) plus an outside-register allowance of US$15M for other robot-model and science-model developers not in the register (for example Latent Labs, Periodic Labs, Absci); the card's own estimate for this part is US$140M-200M-350M (low-central-high). Low and high shares hold the denominator at its central."}},"screening":{"speech-voice":{"levelLow":1,"levelHigh":3},"music":{"levelLow":1,"levelHigh":2},"image-video":{"levelLow":1,"levelHigh":2},"avatar-video":{"levelLow":1,"levelHigh":2},"robotics-science":{"levelLow":1,"levelHigh":2}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"61143fb49129b5d6b081e68a5bb0857882ca4fa537ff9664a7605e9ddb32c3af"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"61143fb49129b5d6b081e68a5bb0857882ca4fa537ff9664a7605e9ddb32c3af"}}},"soundhound":{"schema":"5lc-index-inputs/1","key":"models-specialist-models","entityId":"soundhound","name":"SoundHound AI","kind":"public","commercialActivityVerified":true,"shares":{"speech-voice":{"low":3.1587,"central":6.262,"high":13.1766,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Q2 2026 revenue × 4, share counted","periodEnd":"2026-06-30","source":{"url":"https://www.sec.gov/Archives/edgar/data/0001840856/000184085626000022/soun-20260630.htm","title":"SoundHound AI, Inc. Form 10-Q for the quarter ended June 30, 2026","publisher":"SoundHound AI, Inc. (SEC EDGAR)","publishedDate":"2026-08-10","tier":"T5","accessedDate":"2026-10-23","locator":"Note 4 (Revenue Recognition), three months ended 30 June 2026: revenue by performance obligation (pp. 19-20) hosted services US$32,183K, licensing US$21,370K, professional services US$8,289K, monetization US$55K, total US$61,897K; revenue by service type, a second table (p. 21) service subscriptions US$47,004K, product royalties US$14,838K, monetization US$55K, total US$61,897K (Q2 2025: US$20,918K, US$18,327K, US$3,320K, US$118K = US$42,683K; US$32,057K, US$10,508K, US$118K = US$42,683K)"},"note":"5LayerCake estimate (T5) from a filing (T1). SoundHound's Form 10-Q (filed 10 August 2026) gives Q2 2026 revenue of US$61.9M; × 4 = US$247.6M. Only the part built on its own speech models counts here, because it also sells customer-service and workflow agents (Interactions, acquired 3 September 2025; Amelia). Note 4 cuts the same US$61.9M two ways: by performance obligation (hosted services US$32.2M, licensing US$21.4M, professional services US$8.3M, monetization US$0.06M) and by service type (service subscriptions US$47.0M, product royalties US$14.8M, monetization US$0.06M). Product royalties therefore sit inside licensing and are not an addition to it: licensing US$21.370M − royalties US$14.838M = US$6.532M, and hosted US$32.183M + professional services US$8.289M + that US$6.532M = US$47.004M of subscriptions. The filing does not say this in words; it follows from the two tables adding to the same total, and from licensing being 'on royalty arrangements' (p. 18). Low = product royalties only (voice-enabled cars and devices, US$14.8M × 4 = US$59.4M). Central = licensing, which already contains the royalties, + 25% of hosted services = (21.370 + 0.25 × 32.183) × 4 = US$117.7M. High = all revenue, US$247.6M. The filing gives no revenue split by product line, so the 25% of hosted services is a judgement. Share = the central US$ figure ÷ the part total of US$1,879M. That is the screening total (the register candidates' central figures of US$1,779M plus an outside-register allowance of US$100M for other developers not in the register), held fixed so every company in this sublayer is measured against one total; this company's correction does not change it. The card's own estimate for this part is US$1,050M-1,440M-2,280M (low-central-high). Low and high shares hold the denominator at its central."},"music":{"participates":false,"evidence":{"url":"https://www.sec.gov/Archives/edgar/data/0001840856/000184085626000022/soun-20260630.htm","title":"SoundHound AI, Inc. Form 10-Q for the quarter ended June 30, 2026","publisher":"SoundHound AI, Inc. (SEC EDGAR)","publishedDate":"2026-08-10","tier":"T1","accessedDate":"2026-10-22","locator":"Note 1 (p. 7) describes SoundHound's voice AI platform, Chat AI voice assistant, Houndify and connectivity solutions, and a music app; the filing names no music-generation, image, video, avatar-video or robot model product"}},"image-video":{"participates":false,"evidence":{"url":"https://www.sec.gov/Archives/edgar/data/0001840856/000184085626000022/soun-20260630.htm","title":"SoundHound AI, Inc. Form 10-Q for the quarter ended June 30, 2026","publisher":"SoundHound AI, Inc. (SEC EDGAR)","publishedDate":"2026-08-10","tier":"T1","accessedDate":"2026-10-22","locator":"Note 1 (p. 7) describes SoundHound's voice AI platform, Chat AI voice assistant, Houndify and connectivity solutions, and a music app; the filing names no music-generation, image, video, avatar-video or robot model product"}},"avatar-video":{"participates":false,"evidence":{"url":"https://www.sec.gov/Archives/edgar/data/0001840856/000184085626000022/soun-20260630.htm","title":"SoundHound AI, Inc. Form 10-Q for the quarter ended June 30, 2026","publisher":"SoundHound AI, Inc. (SEC EDGAR)","publishedDate":"2026-08-10","tier":"T1","accessedDate":"2026-10-22","locator":"Note 1 (p. 7) describes SoundHound's voice AI platform, Chat AI voice assistant, Houndify and connectivity solutions, and a music app; the filing names no music-generation, image, video, avatar-video or robot model product"}},"robotics-science":{"participates":false,"evidence":{"url":"https://www.sec.gov/Archives/edgar/data/0001840856/000184085626000022/soun-20260630.htm","title":"SoundHound AI, Inc. Form 10-Q for the quarter ended June 30, 2026","publisher":"SoundHound AI, Inc. (SEC EDGAR)","publishedDate":"2026-08-10","tier":"T1","accessedDate":"2026-10-22","locator":"Note 1 (p. 7) describes SoundHound's voice AI platform, Chat AI voice assistant, Houndify and connectivity solutions, and a music app; the filing names no music-generation, image, video, avatar-video or robot model product"}}},"priorShares":{"speech-voice":{"low":5.0107,"central":11.2329,"high":20.3534,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Q2 2025 revenue × 4, share counted","periodEnd":"2025-06-30","source":{"url":"https://www.sec.gov/Archives/edgar/data/0001840856/000184085626000022/soun-20260630.htm","title":"SoundHound AI, Inc. Form 10-Q for the quarter ended June 30, 2026","publisher":"SoundHound AI, Inc. (SEC EDGAR)","publishedDate":"2026-08-10","tier":"T5","accessedDate":"2026-10-23","locator":"Note 4, Q2 2025 comparative columns: by performance obligation (pp. 19-20) hosted services US$20,918K, licensing US$18,327K, professional services US$3,320K, monetization US$118K, total US$42,683K; by service type (p. 21) service subscriptions US$32,057K, product royalties US$10,508K, monetization US$118K"},"note":"5LayerCake estimate (T5) from the filing's comparative columns (T1), built the same way as the current figure: product royalties sit inside licensing (licensing US$18.327M − royalties US$10.508M = US$7.819M; hosted US$20.918M + professional services US$3.320M + US$7.819M = US$32.057M of subscriptions). Central = (18.327 + 0.25 × 20.918) × 4 = US$94.2M; low = royalties × 4 = US$42.0M; high = total × 4 = US$170.7M. The prior quarter excludes Interactions (acquired 3 September 2025), whose revenue the filing does not separate, so this growth includes acquired revenue. Prior share = the prior US$ figure ÷ the prior part total of US$839M, which is the current part total of US$1,879M ÷ 2.24, the growth of this part over the year: pooled from the four speech companies with a sourced current and prior figure (ElevenLabs US$600M ÷ US$200M, AssemblyAI US$76M ÷ US$26M, SoundHound's counted share US$117.7M ÷ US$94.2M, iFLYTEK's counted share US$83.5M ÷ US$70.8M), US$877.2M ÷ US$391.0M = 2.24. Companies without a sourced prior are assumed to have grown at that rate. Low and high hold the denominator at its central."}},"activityGrowth":{"current":{"low":61897000,"central":61897000,"high":61897000,"unit":"USD","basis":"Total revenue, calendar quarter, US dollars","period":"Q2 2026 (calendar quarter)","periodEnd":"2026-06-30","source":{"url":"https://www.sec.gov/Archives/edgar/data/0001840856/000184085626000022/soun-20260630.htm","title":"SoundHound AI, Inc. Form 10-Q for the quarter ended June 30, 2026","publisher":"SoundHound AI, Inc. (SEC EDGAR)","publishedDate":"2026-08-10","tier":"T1","accessedDate":"2026-10-22","locator":"Condensed Consolidated Statements of Operations: revenue for the three months ended 30 June 2026, US$61,897K"},"note":"All SoundHound revenue (voice AI and the agent products built on it). Includes revenue from acquisitions made after Q2 2025."},"prior":{"low":42683000,"central":42683000,"high":42683000,"unit":"USD","basis":"Total revenue, calendar quarter, US dollars","period":"Q2 2025 (calendar quarter)","periodEnd":"2025-06-30","source":{"url":"https://www.sec.gov/Archives/edgar/data/0001840856/000184085626000022/soun-20260630.htm","title":"SoundHound AI, Inc. Form 10-Q for the quarter ended June 30, 2026","publisher":"SoundHound AI, Inc. (SEC EDGAR)","publishedDate":"2026-08-10","tier":"T1","accessedDate":"2026-10-22","locator":"Condensed Consolidated Statements of Operations: revenue for the three months ended 30 June 2025, US$42,683K"},"note":"Same filing, comparative column. Growth 61.897 ÷ 42.683 = 1.45 times (45%), of which an unquantified part is acquired (Interactions, closed 3 September 2025)."}},"committedExpansion":{"low":0.2423,"central":0.2423,"high":0.2423,"unit":"ratio","basis":"Remaining performance obligations ÷ annualised quarterly revenue (quarter × 4)","period":"Remaining performance obligations at 30 June 2026","periodEnd":"2026-06-30","source":{"url":"https://www.sec.gov/Archives/edgar/data/0001840856/000184085626000022/soun-20260630.htm","title":"SoundHound AI, Inc. Form 10-Q for the quarter ended June 30, 2026","publisher":"SoundHound AI, Inc. (SEC EDGAR)","publishedDate":"2026-08-10","tier":"T1","accessedDate":"2026-10-22","locator":"Note 4, Contract Balances: remaining performance obligations US$60.0M at 30 June 2026, of which US$37.3M within one year, US$20.1M in years 2 to 5 and US$2.6M after 5 years"},"note":"Remaining performance obligations US$60.0M (10-Q, Note 4) ÷ annualised Q2 2026 revenue US$247.6M (61.897 × 4) = 0.2423. Both are the whole company, so the ratio does not depend on how much of the revenue counts as speech. The 10-Q does not quantify SoundHound's wider 'bookings backlog', which includes unsigned volume, so only the contractual figure is used."},"baselineRevenueUsd":{"low":247588000,"central":247588000,"high":247588000,"unit":"USD","basis":"Total revenue, latest calendar quarter × 4, US dollars","period":"Q2 2026 (calendar quarter) × 4","periodEnd":"2026-06-30","source":{"url":"https://www.sec.gov/Archives/edgar/data/0001840856/000184085626000022/soun-20260630.htm","title":"SoundHound AI, Inc. Form 10-Q for the quarter ended June 30, 2026","publisher":"SoundHound AI, Inc. (SEC EDGAR)","publishedDate":"2026-08-10","tier":"T1","accessedDate":"2026-10-22","locator":"Condensed Consolidated Statements of Operations: revenue for the three months ended 30 June 2026, US$61,897K"},"note":"All SoundHound revenue annualised (voice AI and the agents built on it); the share used for scoring counts only part of it."},"screening":{"speech-voice":{"levelLow":1,"levelHigh":3}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"cfdbd0d08924ec1839cb8e31010c6a16355e1d5c4ad7a8544f0f5455011580df"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"cfdbd0d08924ec1839cb8e31010c6a16355e1d5c4ad7a8544f0f5455011580df"}}},"iflytek":{"schema":"5lc-index-inputs/1","key":"models-specialist-models","entityId":"iflytek","name":"iFlytek (speech)","kind":"public","commercialActivityVerified":true,"shares":{"speech-voice":{"low":2.2213,"central":4.4426,"high":9.2553,"unit":"percent","basis":"Calendar-year 2025 revenue in US dollars (RMB converted at the CY2025 average), not the card's annualised run-rate","offBasis":true,"period":"CY2025 revenue, converted at RMB 7.188/US$ (CY2025 average)","periodEnd":"2025-12-31","source":{"url":"https://static.cninfo.com.cn/finalpage/2026-04-29/1225233581.PDF","title":"科大讯飞股份有限公司 2025年年度报告 (iFLYTEK 2025 annual report)","publisher":"iFLYTEK Co., Ltd. (cninfo.com.cn)","publishedDate":"2026-04-29","accessedDate":"2026-10-22","tier":"T5","locator":"PDF p. 11, section (二) AI开放平台: AI平台及授权服务收入12.52亿元 (RMB1.252bn), of which 大模型API及MaaS平台服务收入3.85亿元 (RMB385m, +263%)"},"note":"5LayerCake estimate (T5) from a filing figure (T1). iFLYTEK's 2025 annual report (cninfo, 29 April 2026) gives AI platform and authorization services revenue of RMB1.252bn, of which large-model API and MaaS RMB385m, leaving about RMB0.87bn for speech and other AI capabilities and licences. Speech is assumed to be 35-100% of that remainder: central RMB0.6bn, low RMB0.3bn, high RMB1.25bn (the whole line). Conversion at the CY2025 average of RMB 7.188/US$, Federal Reserve H.10 monthly averages, FRED series EXCHUS, https://fred.stlouisfed.org/data/EXCHUS.txt, last updated 1 October 2026: RMB600M ÷ 7.188 = US$83.5M. Calendar-year revenue, so flagged off-basis. Cross-check from the 2026 interim report (T1, PDF pp. 29-30): the open platform segment earned RMB3,704.7M in H1 2026, up 36.0% on RMB2,723.9M, and large-model API and MaaS revenue grew about 70%, so the current run-rate is probably higher than CY2025. Share = the central US$ figure ÷ the part total of US$1,879M. That is the screening total (the register candidates' central figures of US$1,779M plus an outside-register allowance of US$100M for other developers not in the register), held fixed so every company in this sublayer is measured against one total; the card's own estimate for this part is US$1,050M-1,440M-2,280M (low-central-high). Low and high shares hold the denominator at its central."},"music":{"low":0,"central":0.2602,"high":1.5613,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"CY2025 estimate","periodEnd":"2025-12-31","source":{"url":"https://static.cninfo.com.cn/finalpage/2026-04-29/1225233581.PDF","title":"科大讯飞股份有限公司 2025年年度报告 (iFLYTEK 2025 annual report)","publisher":"iFLYTEK Co., Ltd. (cninfo.com.cn)","publishedDate":"2026-04-29","accessedDate":"2026-10-22","tier":"T5","locator":"PDF p. 11, section (二) AI开放平台: AI平台及授权服务收入12.52亿元 (RMB1.252bn), of which 大模型API及MaaS平台服务收入3.85亿元 (RMB385m, +263%)"},"note":"5LayerCake estimate (T5). The annual report describes iFLYTEK Music as a song-distribution business (PDF p. 27) and in-car sound features; no AI music-model revenue is stated. Allowance only. Share = the central US$ figure ÷ the part total of US$384M, which is the register candidates' central figures (US$364M) plus an outside-register allowance of US$20M for other music-model developers not in the register (for example Producer.ai, AIVA, Soundraw); the card's own estimate for this part is US$315M-336M-500M (low-central-high). Low and high shares hold the denominator at its central."},"image-video":{"low":0.0266,"central":0.133,"high":0.7982,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"CY2025 estimate","periodEnd":"2025-12-31","source":{"url":"https://static.cninfo.com.cn/finalpage/2026-04-29/1225233581.PDF","title":"科大讯飞股份有限公司 2025年年度报告 (iFLYTEK 2025 annual report)","publisher":"iFLYTEK Co., Ltd. (cninfo.com.cn)","publishedDate":"2026-04-29","accessedDate":"2026-10-22","tier":"T5","locator":"PDF p. 11, section (二) AI开放平台: AI平台及授权服务收入12.52亿元 (RMB1.252bn), of which 大模型API及MaaS平台服务收入3.85亿元 (RMB385m, +263%)"},"note":"5LayerCake estimate (T5). The annual report lists video generation among the scenarios of the Spark model matrix (PDF p. 11) and gives no revenue. Allowance only. Share = the central US$ figure ÷ the part total of US$3,758M, which is the register candidates' central figures (US$3,608M) plus an outside-register allowance of US$150M for other image and video developers not in the register (for example Higgsfield, Freepik, Tripo, Genmo); the card's own estimate for this part is US$2,140M-3,590M-5,660M (low-central-high). Low and high shares hold the denominator at its central."},"avatar-video":{"low":0.2011,"central":1.0054,"high":5.0271,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"CY2025 estimate","periodEnd":"2025-12-31","source":{"url":"https://static.cninfo.com.cn/finalpage/2026-04-29/1225233581.PDF","title":"科大讯飞股份有限公司 2025年年度报告 (iFLYTEK 2025 annual report)","publisher":"iFLYTEK Co., Ltd. (cninfo.com.cn)","publishedDate":"2026-04-29","accessedDate":"2026-10-22","tier":"T5","locator":"PDF p. 11, section (二) AI开放平台: AI平台及授权服务收入12.52亿元 (RMB1.252bn), of which 大模型API及MaaS平台服务收入3.85亿元 (RMB385m, +263%)"},"note":"5LayerCake estimate (T5). The annual report describes a virtual human platform with 100,000+ enterprise digital twins and 5 million voice models (PDF p. 12) and gives no revenue. Allowance only; learning apps count in Applications. Share = the central US$ figure ÷ the part total of US$497M, which is the register candidates' central figures (US$467M) plus an outside-register allowance of US$30M for other avatar-video developers not in the register (for example Akool, Argil, Creatify); the card's own estimate for this part is US$400M-450M-550M (low-central-high). Low and high shares hold the denominator at its central."},"robotics-science":{"low":0,"central":1.2837,"high":8.558,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"CY2025 estimate","periodEnd":"2025-12-31","source":{"url":"https://static.cninfo.com.cn/finalpage/2026-04-29/1225233581.PDF","title":"科大讯飞股份有限公司 2025年年度报告 (iFLYTEK 2025 annual report)","publisher":"iFLYTEK Co., Ltd. (cninfo.com.cn)","publishedDate":"2026-04-29","accessedDate":"2026-10-22","tier":"T5","locator":"PDF p. 11, section (二) AI开放平台: AI平台及授权服务收入12.52亿元 (RMB1.252bn), of which 大模型API及MaaS平台服务收入3.85亿元 (RMB385m, +263%)"},"note":"5LayerCake estimate (T5). The annual report describes a robot 'super-brain' platform used by most domestic robot makers (PDF p. 12) and gives no revenue; it is mainly voice and interaction modules. Allowance only. Share = the central US$ figure ÷ the part total of US$234M, which is the register candidates' central figures (US$219M) plus an outside-register allowance of US$15M for other robot-model and science-model developers not in the register (for example Latent Labs, Periodic Labs, Absci); the card's own estimate for this part is US$140M-200M-350M (low-central-high). Low and high shares hold the denominator at its central."}},"priorShares":{"speech-voice":{"low":4.2218,"central":8.4437,"high":17.5908,"unit":"percent","basis":"Calendar-year 2024 revenue in US dollars (RMB converted at the CY2024 average), not the card's annualised run-rate","offBasis":true,"period":"CY2024 estimate, converted at RMB 7.196/US$ (CY2024 average)","periodEnd":"2024-12-31","source":{"url":"https://static.cninfo.com.cn/finalpage/2026-04-29/1225233581.PDF","title":"科大讯飞股份有限公司 2025年年度报告 (iFLYTEK 2025 annual report)","publisher":"iFLYTEK Co., Ltd. (cninfo.com.cn)","publishedDate":"2026-04-29","accessedDate":"2026-10-22","tier":"T5","locator":"PDF p. 34, revenue by product: 开放平台 (open platform) RMB6,088.3M in 2025 against RMB5,171.9M in 2024 (+17.7%)"},"note":"5LayerCake estimate (T5) from filing figures (T1). The 2024 value of the AI platform and authorization services line is not disclosed, so the current speech estimate is scaled back by the growth of the whole open platform segment (RMB6,088.3M ÷ RMB5,171.9M = 1.1772, 2025 annual report PDF p. 34): RMB600M ÷ 1.1772 = RMB510M; ÷ the CY2024 average of RMB 7.196/US$ (Federal Reserve H.10 monthly averages, FRED series EXCHUS, https://fred.stlouisfed.org/data/EXCHUS.txt, last updated 1 October 2026) = US$70.8M. The segment includes lines that grew slower or faster than speech, so this is a proxy. Prior share = the prior US$ figure ÷ the prior part total of US$839M, which is the current part total of US$1,879M ÷ 2.24, the growth of this part over the year: pooled from the four speech companies with a sourced current and prior figure (ElevenLabs US$600M ÷ US$200M, AssemblyAI US$76M ÷ US$26M, SoundHound's counted share US$117.7M ÷ US$94.2M, iFLYTEK's counted share US$83.5M ÷ US$70.8M), US$877.2M ÷ US$391.0M = 2.24. Companies without a sourced prior are assumed to have grown at that rate. Low and high hold the denominator at its central."},"music":{"low":0,"central":0.2602,"high":1.5613,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Held at the current share (prior period 2024-12-31)","periodEnd":"2024-12-31","source":{"url":"https://static.cninfo.com.cn/finalpage/2026-04-29/1225233581.PDF","title":"科大讯飞股份有限公司 2025年年度报告 (iFLYTEK 2025 annual report)","publisher":"iFLYTEK Co., Ltd. (cninfo.com.cn)","publishedDate":"2026-04-29","accessedDate":"2026-10-22","tier":"T5","locator":"PDF p. 11, section (二) AI开放平台: AI平台及授权服务收入12.52亿元 (RMB1.252bn), of which 大模型API及MaaS平台服务收入3.85亿元 (RMB385m, +263%)"},"note":"5LayerCake assumption (T5): no figure for the earlier period exists for this undisclosed allowance, so the share is held at the current share (the company is assumed to have grown with this part of the market)."},"image-video":{"low":0.0266,"central":0.133,"high":0.7982,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Held at the current share (prior period 2024-12-31)","periodEnd":"2024-12-31","source":{"url":"https://static.cninfo.com.cn/finalpage/2026-04-29/1225233581.PDF","title":"科大讯飞股份有限公司 2025年年度报告 (iFLYTEK 2025 annual report)","publisher":"iFLYTEK Co., Ltd. (cninfo.com.cn)","publishedDate":"2026-04-29","accessedDate":"2026-10-22","tier":"T5","locator":"PDF p. 11, section (二) AI开放平台: AI平台及授权服务收入12.52亿元 (RMB1.252bn), of which 大模型API及MaaS平台服务收入3.85亿元 (RMB385m, +263%)"},"note":"5LayerCake assumption (T5): no figure for the earlier period exists for this undisclosed allowance, so the share is held at the current share (the company is assumed to have grown with this part of the market)."},"avatar-video":{"low":0.2011,"central":1.0054,"high":5.0271,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Held at the current share (prior period 2024-12-31)","periodEnd":"2024-12-31","source":{"url":"https://static.cninfo.com.cn/finalpage/2026-04-29/1225233581.PDF","title":"科大讯飞股份有限公司 2025年年度报告 (iFLYTEK 2025 annual report)","publisher":"iFLYTEK Co., Ltd. (cninfo.com.cn)","publishedDate":"2026-04-29","accessedDate":"2026-10-22","tier":"T5","locator":"PDF p. 11, section (二) AI开放平台: AI平台及授权服务收入12.52亿元 (RMB1.252bn), of which 大模型API及MaaS平台服务收入3.85亿元 (RMB385m, +263%)"},"note":"5LayerCake assumption (T5): no figure for the earlier period exists for this undisclosed allowance, so the share is held at the current share (the company is assumed to have grown with this part of the market)."},"robotics-science":{"low":0,"central":1.2837,"high":8.558,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Held at the current share (prior period 2024-12-31)","periodEnd":"2024-12-31","source":{"url":"https://static.cninfo.com.cn/finalpage/2026-04-29/1225233581.PDF","title":"科大讯飞股份有限公司 2025年年度报告 (iFLYTEK 2025 annual report)","publisher":"iFLYTEK Co., Ltd. (cninfo.com.cn)","publishedDate":"2026-04-29","accessedDate":"2026-10-22","tier":"T5","locator":"PDF p. 11, section (二) AI开放平台: AI平台及授权服务收入12.52亿元 (RMB1.252bn), of which 大模型API及MaaS平台服务收入3.85亿元 (RMB385m, +263%)"},"note":"5LayerCake assumption (T5): no figure for the earlier period exists for this undisclosed allowance, so the share is held at the current share (the company is assumed to have grown with this part of the market)."}},"activityGrowth":{"current":{"low":3704729699,"central":3704729699,"high":3704729699,"unit":"CNY","basis":"Open platform segment revenue, half-year, RMB (proxy for iFLYTEK's AI platform and speech revenue)","period":"H1 2026 (half-year)","periodEnd":"2026-06-30","source":{"url":"http://static.cninfo.com.cn/finalpage/2026-08-21/1225485550.PDF","title":"科大讯飞股份有限公司2026年半年度报告 (iFLYTEK Co., Ltd. 2026 interim report)","publisher":"iFLYTEK Co., Ltd. (cninfo.com.cn)","publishedDate":"2026-08","tier":"T1","accessedDate":"2026-10-22","locator":"PDF pp. 29-30, revenue by product: 开放平台 RMB3,704,729,699.82 (+36.01%)"},"note":"Segment revenue, not the sublayer's speech revenue alone: the segment also holds API, MaaS, authorization and other platform lines. Used as a labelled proxy because iFLYTEK does not report its speech revenue."},"prior":{"low":2723932885,"central":2723932885,"high":2723932885,"unit":"CNY","basis":"Open platform segment revenue, half-year, RMB (proxy for iFLYTEK's AI platform and speech revenue)","period":"H1 2025 (half-year)","periodEnd":"2025-06-30","source":{"url":"http://static.cninfo.com.cn/finalpage/2026-08-21/1225485550.PDF","title":"科大讯飞股份有限公司2026年半年度报告 (iFLYTEK Co., Ltd. 2026 interim report)","publisher":"iFLYTEK Co., Ltd. (cninfo.com.cn)","publishedDate":"2026-08","tier":"T1","accessedDate":"2026-10-22","locator":"PDF pp. 29-30, revenue by product, prior-period column: 开放平台 RMB2,723,932,884.53"},"note":"Same table, comparative column. Growth 3,704.7 ÷ 2,723.9 = 1.36 times."}},"screening":{"speech-voice":{"levelLow":1,"levelHigh":3},"music":{"levelLow":1,"levelHigh":2},"image-video":{"levelLow":1,"levelHigh":2},"avatar-video":{"levelLow":1,"levelHigh":2},"robotics-science":{"levelLow":1,"levelHigh":3}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"d344b6c4f86153be78f36c3432ffa843981102b166c855743bf017567aebceff"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"d344b6c4f86153be78f36c3432ffa843981102b166c855743bf017567aebceff"}}},"amazon-agi":{"schema":"5lc-index-inputs/1","key":"models-specialist-models","entityId":"amazon-agi","name":"Amazon Nova media models, Polly and Transcribe","kind":"unit","parentId":"amazon","commercialActivityVerified":true,"shares":{"speech-voice":{"low":2.6614,"central":7.452,"high":18.6299,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Estimate for October 2026","periodEnd":"2026-10-22","source":{"url":"https://aws.amazon.com/polly/pricing/","title":"Amazon Polly pricing","publisher":"Amazon Web Services","publishedDate":"undated","accessedDate":"2026-10-22","tier":"T5","locator":"Per-character prices: Standard $4, Neural $16, Long-Form $100, Generative $30 per million characters"},"note":"5LayerCake estimate (T5). Amazon sells Polly ($4-100 per million characters), Transcribe and Nova Sonic; no revenue is disclosed. Share of the card's US$600M large-company speech allowance: 140. Unit figure, not Amazon's total. Share = the central US$ figure ÷ the part total of US$1,879M, which is the register candidates' central figures (US$1,779M) plus an outside-register allowance of US$100M for other speech developers not in the register (for example Gladia, Soniox, Fish Audio, Murf, WellSaid, Rev); the card's own estimate for this part is US$1,050M-1,440M-2,280M (low-central-high). Low and high shares hold the denominator at its central."},"music":{"low":0,"central":0.1301,"high":1.3011,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Estimate for October 2026","periodEnd":"2026-10-22","source":{"url":"https://aws.amazon.com/blogs/aws/introducing-amazon-nova-frontier-intelligence-and-industry-leading-price-performance/","title":"Introducing Amazon Nova foundation models: Frontier intelligence and industry leading price performance","publisher":"Amazon Web Services (AWS News Blog)","publishedDate":"2024-12-03","accessedDate":"2026-10-23","tier":"T5","locator":"Sections 'Creative content generation' and the model list: Nova Micro, Lite, Pro and Premier (text and multimodal understanding) plus 'two creative content generation models', Nova Canvas (image) and Nova Reel (video); the Nova models are 'available exclusively in Amazon Bedrock'; no music, avatar-video or robot model is named (post of 3 December 2024, updated 23 April 2025)"},"note":"5LayerCake judgment (T5), not a measured figure. AWS's announcement of the Nova family (3 December 2024, opened) lists the text and multimodal models and two creative content generation models, Nova Canvas (image) and Nova Reel (video), and names no music-generation model; the AWS Nova product page (https://aws.amazon.com/ai/generative-ai/nova/, opened 23 October 2026) lists Nova 2 Lite, Nova 2 Sonic (speech-to-speech), Nova Act and Nova Forge and also names no music, avatar-video or robot model. Absence from those pages does not prove Amazon earns nothing here, so a token allowance is kept instead of a non-participation claim: central US$0.5M, high US$5M (10 times central), low zero. Share = the central US$ figure ÷ the part total of US$384M, which is the register candidates' central figures (US$364M) plus an outside-register allowance of US$20M for other music-model developers not in the register (for example Producer.ai, AIVA, Soundraw); the card's own estimate for this part is US$315M-336M-500M (low-central-high). Low and high shares hold the denominator at its central."},"image-video":{"low":0.133,"central":0.6652,"high":2.1286,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Estimate for October 2026","periodEnd":"2026-10-22","source":{"url":"https://aws.amazon.com/blogs/aws/introducing-amazon-nova-frontier-intelligence-and-industry-leading-price-performance/","title":"Introducing Amazon Nova foundation models: Frontier intelligence and industry leading price performance","publisher":"Amazon Web Services (AWS News Blog)","publishedDate":"2024-12-03","accessedDate":"2026-10-23","tier":"T5","locator":"Sections 'Creative content generation' and the model list: Nova Micro, Lite, Pro and Premier (text and multimodal understanding) plus 'two creative content generation models', Nova Canvas (image) and Nova Reel (video); the Nova models are 'available exclusively in Amazon Bedrock'; no music, avatar-video or robot model is named (post of 3 December 2024, updated 23 April 2025)"},"note":"5LayerCake judgment (T5) anchored on a company page. Nova Canvas (images) and Nova Reel (video) are Amazon's own models and are sold 'exclusively in Amazon Bedrock' (AWS announcement of 3 December 2024, opened); AWS gives no usage or revenue for either. Share of the card's US$800M central allowance (range US$300-1,600M) for the image and video models of Google, OpenAI, Microsoft and Amazon: Amazon is given about 3% (0.03 × US$800M = US$24M, rounded to US$25M central), far below Google's and OpenAI's shares of the same allowance in this sublayer (about 50% and 31%), because the two models are sold only through one cloud marketplace and no usage figures were found. Low US$5M (about 1.7% of the US$300M low allowance); high US$80M (5% of the US$1,600M high allowance). Unit figure, not Amazon's total. Share = the central US$ figure ÷ the part total of US$3,758M, which is the register candidates' central figures (US$3,608M) plus an outside-register allowance of US$150M for other image and video developers not in the register (for example Higgsfield, Freepik, Tripo, Genmo); the card's own estimate for this part is US$2,140M-3,590M-5,660M (low-central-high). Low and high shares hold the denominator at its central."},"avatar-video":{"low":0,"central":0.1005,"high":1.0054,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Estimate for October 2026","periodEnd":"2026-10-22","source":{"url":"https://aws.amazon.com/blogs/aws/introducing-amazon-nova-frontier-intelligence-and-industry-leading-price-performance/","title":"Introducing Amazon Nova foundation models: Frontier intelligence and industry leading price performance","publisher":"Amazon Web Services (AWS News Blog)","publishedDate":"2024-12-03","accessedDate":"2026-10-23","tier":"T5","locator":"Sections 'Creative content generation' and the model list: Nova Micro, Lite, Pro and Premier (text and multimodal understanding) plus 'two creative content generation models', Nova Canvas (image) and Nova Reel (video); the Nova models are 'available exclusively in Amazon Bedrock'; no music, avatar-video or robot model is named (post of 3 December 2024, updated 23 April 2025)"},"note":"5LayerCake judgment (T5), not a measured figure. AWS's announcement of the Nova family (3 December 2024, opened) lists the text and multimodal models and two creative content generation models, Nova Canvas (image) and Nova Reel (video), and names no avatar-video model; the AWS Nova product page (https://aws.amazon.com/ai/generative-ai/nova/, opened 23 October 2026) lists Nova 2 Lite, Nova 2 Sonic (speech-to-speech), Nova Act and Nova Forge and also names no music, avatar-video or robot model. Absence from those pages does not prove Amazon earns nothing here, so a token allowance is kept instead of a non-participation claim: central US$0.5M, high US$5M (10 times central), low zero. Share = the central US$ figure ÷ the part total of US$497M, which is the register candidates' central figures (US$467M) plus an outside-register allowance of US$30M for other avatar-video developers not in the register (for example Akool, Argil, Creatify); the card's own estimate for this part is US$400M-450M-550M (low-central-high). Low and high shares hold the denominator at its central."},"robotics-science":{"low":0,"central":0.1284,"high":1.2837,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Estimate for October 2026","periodEnd":"2026-10-22","source":{"url":"https://aws.amazon.com/blogs/aws/introducing-amazon-nova-frontier-intelligence-and-industry-leading-price-performance/","title":"Introducing Amazon Nova foundation models: Frontier intelligence and industry leading price performance","publisher":"Amazon Web Services (AWS News Blog)","publishedDate":"2024-12-03","accessedDate":"2026-10-23","tier":"T5","locator":"Sections 'Creative content generation' and the model list: Nova Micro, Lite, Pro and Premier (text and multimodal understanding) plus 'two creative content generation models', Nova Canvas (image) and Nova Reel (video); the Nova models are 'available exclusively in Amazon Bedrock'; no music, avatar-video or robot model is named (post of 3 December 2024, updated 23 April 2025)"},"note":"5LayerCake judgment (T5), not a measured figure. AWS's announcement of the Nova family (3 December 2024, opened) lists the text and multimodal models and two creative content generation models, Nova Canvas (image) and Nova Reel (video), and names no robot or science model; the AWS Nova product page (https://aws.amazon.com/ai/generative-ai/nova/, opened 23 October 2026) lists Nova 2 Lite, Nova 2 Sonic (speech-to-speech), Nova Act and Nova Forge and also names no music, avatar-video or robot model. Absence from those pages does not prove Amazon earns nothing here, so a token allowance is kept instead of a non-participation claim: central US$0.3M, high US$3M (10 times central), low zero. Share = the central US$ figure ÷ the part total of US$234M, which is the register candidates' central figures (US$219M) plus an outside-register allowance of US$15M for other robot-model and science-model developers not in the register (for example Latent Labs, Periodic Labs, Absci); the card's own estimate for this part is US$140M-200M-350M (low-central-high). Low and high shares hold the denominator at its central."}},"screening":{"speech-voice":{"levelLow":1,"levelHigh":2},"music":{"levelLow":1,"levelHigh":2},"image-video":{"levelLow":1,"levelHigh":2},"avatar-video":{"levelLow":1,"levelHigh":2},"robotics-science":{"levelLow":1,"levelHigh":2}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"7a4dd1009209153c89298f7542c6fb8c1bcbb5fce4ce39bcb3ee64cb93fb05e3"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"7a4dd1009209153c89298f7542c6fb8c1bcbb5fce4ce39bcb3ee64cb93fb05e3"}}},"black-forest-labs":{"schema":"5lc-index-inputs/1","key":"models-specialist-models","entityId":"black-forest-labs","name":"Black Forest Labs","kind":"private","commercialActivityVerified":true,"shares":{"speech-voice":{"participates":false,"evidence":{"url":"https://sacra.com/c/black-forest-labs/","title":"Black Forest Labs (Sacra company profile)","publisher":"Sacra","publishedDate":"undated","accessedDate":"2026-10-22","tier":"T4","locator":"Sacra describes Black Forest Labs as an image-model company (FLUX) selling to enterprises and platforms; no speech, music, avatar or robot models."}},"music":{"participates":false,"evidence":{"url":"https://sacra.com/c/black-forest-labs/","title":"Black Forest Labs (Sacra company profile)","publisher":"Sacra","publishedDate":"undated","accessedDate":"2026-10-22","tier":"T4","locator":"Sacra describes Black Forest Labs as an image-model company (FLUX) selling to enterprises and platforms; no speech, music, avatar or robot models."}},"image-video":{"low":2.5623,"central":2.5623,"high":2.5623,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"ARR, August 2025","periodEnd":"2025-08-31","source":{"url":"https://sacra.com/c/black-forest-labs/","title":"Black Forest Labs (Sacra company profile)","publisher":"Sacra","publishedDate":"undated","accessedDate":"2026-10-22","tier":"T4","locator":"Sacra estimates about $96M annualized revenue as of August 2025; $140M Meta contract (Sept 2025); about $300M contract value across partners; $300M Series B at $3.25B (Dec 2025)"},"note":"Market-research estimate by Sacra: about $96M annualized revenue as of August 2025. It is more than 12 months old, so the engine widens it. Later evidence points the same way or higher: a $140M multi-year Meta contract (Sept 2025), about $300M contract value across partners including Adobe, Canva and Snap, and a $300M Series B at $3.25B (Dec 2025). Contract value is not recognised revenue. Share = the central US$ figure ÷ the part total of US$3,758M, which is the register candidates' central figures (US$3,608M) plus an outside-register allowance of US$150M for other image and video developers not in the register (for example Higgsfield, Freepik, Tripo, Genmo); the card's own estimate for this part is US$2,140M-3,590M-5,660M (low-central-high). Low and high shares hold the denominator at its central."},"avatar-video":{"participates":false,"evidence":{"url":"https://sacra.com/c/black-forest-labs/","title":"Black Forest Labs (Sacra company profile)","publisher":"Sacra","publishedDate":"undated","accessedDate":"2026-10-22","tier":"T4","locator":"Sacra describes Black Forest Labs as an image-model company (FLUX) selling to enterprises and platforms; no speech, music, avatar or robot models."}},"robotics-science":{"participates":false,"evidence":{"url":"https://sacra.com/c/black-forest-labs/","title":"Black Forest Labs (Sacra company profile)","publisher":"Sacra","publishedDate":"undated","accessedDate":"2026-10-22","tier":"T4","locator":"Sacra describes Black Forest Labs as an image-model company (FLUX) selling to enterprises and platforms; no speech, music, avatar or robot models."}}},"baselineRevenueUsd":{"low":96000000,"central":96000000,"high":96000000,"unit":"USD","basis":"Annualised revenue of Black Forest Labs, US dollars (Sacra estimate)","period":"Annualised revenue, August 2025 (Sacra estimate)","periodEnd":"2025-08-31","source":{"url":"https://sacra.com/c/black-forest-labs/","title":"Black Forest Labs (Sacra company profile)","publisher":"Sacra","publishedDate":"undated","accessedDate":"2026-10-22","tier":"T4","locator":"Sacra estimates Black Forest Labs hit US$96M in annualised revenue as of August 2025"},"note":"Sacra estimate (T4), no method stated; more than 12 months old, so the engine widens it. No newer revenue figure was found in a search of 2026 coverage."},"screening":{"image-video":{"levelLow":1,"levelHigh":3}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"bc921da6a2fa60471413f893684ee2ee3b8f102be33414b48cf0450588ba886f"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"bc921da6a2fa60471413f893684ee2ee3b8f102be33414b48cf0450588ba886f"}}},"field-ai":{"schema":"5lc-index-inputs/1","key":"models-specialist-models","entityId":"field-ai","name":"Field AI","kind":"private","commercialActivityVerified":true,"shares":{"speech-voice":{"participates":false,"evidence":{"url":"https://thenextweb.com/news/fieldai-10b-europe-robot-brains","title":"Robot software startup FieldAI is set to raise $700M at a $10B valuation","publisher":"The Next Web","publishedDate":"2026-10-02","tier":"T3","locator":"The Next Web describes FieldAI as selling robot 'brains' (foundation models for robots) to more than 30 customers; it names no speech, music, image, video or avatar-video product","accessedDate":"2026-10-22"}},"music":{"participates":false,"evidence":{"url":"https://thenextweb.com/news/fieldai-10b-europe-robot-brains","title":"Robot software startup FieldAI is set to raise $700M at a $10B valuation","publisher":"The Next Web","publishedDate":"2026-10-02","tier":"T3","locator":"The Next Web describes FieldAI as selling robot 'brains' (foundation models for robots) to more than 30 customers; it names no speech, music, image, video or avatar-video product","accessedDate":"2026-10-22"}},"image-video":{"participates":false,"evidence":{"url":"https://thenextweb.com/news/fieldai-10b-europe-robot-brains","title":"Robot software startup FieldAI is set to raise $700M at a $10B valuation","publisher":"The Next Web","publishedDate":"2026-10-02","tier":"T3","locator":"The Next Web describes FieldAI as selling robot 'brains' (foundation models for robots) to more than 30 customers; it names no speech, music, image, video or avatar-video product","accessedDate":"2026-10-22"}},"avatar-video":{"participates":false,"evidence":{"url":"https://thenextweb.com/news/fieldai-10b-europe-robot-brains","title":"Robot software startup FieldAI is set to raise $700M at a $10B valuation","publisher":"The Next Web","publishedDate":"2026-10-02","tier":"T3","locator":"The Next Web describes FieldAI as selling robot 'brains' (foundation models for robots) to more than 30 customers; it names no speech, music, image, video or avatar-video product","accessedDate":"2026-10-22"}},"robotics-science":{"low":6.3462,"central":19.0385,"high":42.6923,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Annual run-rate, estimate for October 2026","periodEnd":"2026-10-02","source":{"url":"https://thenextweb.com/news/fieldai-10b-europe-robot-brains","title":"Robot software startup FieldAI is set to raise $700M at a $10B valuation","publisher":"The Next Web","publishedDate":"2026-10-02","tier":"T5","locator":"\"Revenue and customer contracts have passed $135M across more than 30 customers, up at least $35M since June.\"","accessedDate":"2026-10-22"},"note":"5LayerCake estimate (T5) from a press report (T3). The Next Web (2 October 2026) reports that FieldAI's revenue and customer contracts have passed US$135M across more than 30 customers, up at least US$35M since June; it does not split recognised revenue from signed contracts or name the source. Annual run-rate assumed at 11-74% of the contracted total (central 33%): 135 × 0.11 / 0.33 / 0.74 = US$14.9M / 44.6M / 99.9M. Share = the central US$ figure ÷ the part total of US$234M. That is the screening total (the register candidates' central figures of US$219M plus an outside-register allowance of US$15M for other developers not in the register), held fixed so every company in this sublayer is measured against one total; the card's own estimate for this part is US$140M-200M-350M (low-central-high). Low and high shares hold the denominator at its central."}},"screening":{"robotics-science":{"levelLow":1,"levelHigh":3}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"290f8ec4bb172b6cdcf98526ac48b06a188b1235ab5154208991ade165bb34e0"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"290f8ec4bb172b6cdcf98526ac48b06a188b1235ab5154208991ade165bb34e0"}}},"deepgram":{"schema":"5lc-index-inputs/1","key":"models-specialist-models","entityId":"deepgram","name":"Deepgram","kind":"private","commercialActivityVerified":true,"shares":{"speech-voice":{"low":5.322,"central":5.322,"high":5.322,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Annualized revenue, August 2026","periodEnd":"2026-08-31","source":{"url":"https://sacra.com/c/deepgram/","title":"Deepgram (Sacra company profile)","publisher":"Sacra","publishedDate":"undated","accessedDate":"2026-10-22","tier":"T4","locator":"Sacra estimates annualized revenue of $13M (end-2023), $28M (end-2024), $60M (end-2025) and $100M (August 2026)"},"note":"Market-research estimate by Sacra: Deepgram hit US$100M of annualized revenue in August 2026 (path US$13M end-2023, US$28M end-2024, US$60M end-2025; the page states no method and does not use the term ARR). TechCrunch (13 January 2026, https://techcrunch.com/2026/01/13/deepgram-raises-130m-at-1-3b-valuation-and-buys-a-yc-ai-startup) reports 1,300+ organisations using Deepgram's voice AI, a US$130M Series C at a US$1.3B valuation and that the company was cash-flow positive in 2025, but no revenue figure. Secondary sites repeat a company claim of US$100M ARR that was not found on a primary page. Share = the central US$ figure ÷ the part total of US$1,879M. That is the screening total (the register candidates' central figures of US$1,779M plus an outside-register allowance of US$100M for other developers not in the register), held fixed so every company in this sublayer is measured against one total; the card's own estimate for this part is US$1,050M-1,440M-2,280M (low-central-high). Low and high shares hold the denominator at its central."},"music":{"participates":false,"evidence":{"url":"https://techcrunch.com/2026/01/13/deepgram-raises-130m-at-1-3b-valuation-and-buys-a-yc-ai-startup","title":"Deepgram raises $130M at $1.3B valuation and buys a YC AI startup","publisher":"TechCrunch","publishedDate":"2026-01-13","tier":"T3","locator":"Describes Deepgram's speech-to-text, text-to-speech and conversational voice-agent models and 1,300+ organisations; names no music, image, video, avatar or robot product","accessedDate":"2026-10-22"}},"image-video":{"participates":false,"evidence":{"url":"https://techcrunch.com/2026/01/13/deepgram-raises-130m-at-1-3b-valuation-and-buys-a-yc-ai-startup","title":"Deepgram raises $130M at $1.3B valuation and buys a YC AI startup","publisher":"TechCrunch","publishedDate":"2026-01-13","tier":"T3","locator":"Describes Deepgram's speech-to-text, text-to-speech and conversational voice-agent models and 1,300+ organisations; names no music, image, video, avatar or robot product","accessedDate":"2026-10-22"}},"avatar-video":{"participates":false,"evidence":{"url":"https://techcrunch.com/2026/01/13/deepgram-raises-130m-at-1-3b-valuation-and-buys-a-yc-ai-startup","title":"Deepgram raises $130M at $1.3B valuation and buys a YC AI startup","publisher":"TechCrunch","publishedDate":"2026-01-13","tier":"T3","locator":"Describes Deepgram's speech-to-text, text-to-speech and conversational voice-agent models and 1,300+ organisations; names no music, image, video, avatar or robot product","accessedDate":"2026-10-22"}},"robotics-science":{"participates":false,"evidence":{"url":"https://techcrunch.com/2026/01/13/deepgram-raises-130m-at-1-3b-valuation-and-buys-a-yc-ai-startup","title":"Deepgram raises $130M at $1.3B valuation and buys a YC AI startup","publisher":"TechCrunch","publishedDate":"2026-01-13","tier":"T3","locator":"Describes Deepgram's speech-to-text, text-to-speech and conversational voice-agent models and 1,300+ organisations; names no music, image, video, avatar or robot product","accessedDate":"2026-10-22"}}},"baselineRevenueUsd":{"low":100000000,"central":100000000,"high":100000000,"unit":"USD","basis":"Annualised revenue of Deepgram, US dollars (Sacra estimate)","period":"Annualised revenue, August 2026 (Sacra estimate)","periodEnd":"2026-08-31","source":{"url":"https://sacra.com/c/deepgram/","title":"Deepgram (Sacra company profile)","publisher":"Sacra","publishedDate":"undated","accessedDate":"2026-10-22","tier":"T4","locator":"Sacra estimates Deepgram hit US$100M in annualised revenue in August 2026, up from US$60M at the end of 2025"},"note":"Sacra estimate (T4), no method stated. It is the same figure as the speech share."},"screening":{"speech-voice":{"levelLow":1,"levelHigh":2}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"046fa3feae4f8652ca341f2c5ffcb4281944a1c3a8fd615051fee2e93347660e"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"046fa3feae4f8652ca341f2c5ffcb4281944a1c3a8fd615051fee2e93347660e"}}},"topaz-labs":{"schema":"5lc-index-inputs/1","key":"models-specialist-models","entityId":"topaz-labs","name":"Topaz Labs","kind":"private","commercialActivityVerified":true,"shares":{"speech-voice":{"participates":false,"evidence":{"url":"https://news.adobe.com/news/2026/06/adobe-to-acquire-topaz-labs","title":"Adobe to Acquire Topaz Labs","publisher":"Adobe","publishedDate":"2026-06-25","accessedDate":"2026-10-22","tier":"T2","locator":"Adobe's release lists Topaz Photo, Video, Gigapixel, Astra and Bloom (image and video enhancement); no speech, music, avatar or robot models."}},"music":{"participates":false,"evidence":{"url":"https://news.adobe.com/news/2026/06/adobe-to-acquire-topaz-labs","title":"Adobe to Acquire Topaz Labs","publisher":"Adobe","publishedDate":"2026-06-25","accessedDate":"2026-10-22","tier":"T2","locator":"Adobe's release lists Topaz Photo, Video, Gigapixel, Astra and Bloom (image and video enhancement); no speech, music, avatar or robot models."}},"image-video":{"low":0.2661,"central":1.1973,"high":3.1929,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Estimate for October 2026","periodEnd":"2026-10-22","source":{"url":"https://news.adobe.com/news/2026/06/adobe-to-acquire-topaz-labs","title":"Adobe to Acquire Topaz Labs","publisher":"Adobe","publishedDate":"2026-06-25","accessedDate":"2026-10-22","tier":"T5","locator":"Topaz Photo, Video, Gigapixel, Astra and Bloom used by millions of customers; products stay available as standalone offerings; revenue not disclosed"},"note":"5LayerCake estimate (T5). Adobe (25 June 2026) agreed to acquire Topaz Labs; its products are used by 'millions of customers' and stay on sale as standalone software; revenue and price are not disclosed. Allowance sized from a consumer-software company with a user base in the millions. Counts only its own enhancement models sold stand-alone. Share = the central US$ figure ÷ the part total of US$3,758M, which is the register candidates' central figures (US$3,608M) plus an outside-register allowance of US$150M for other image and video developers not in the register (for example Higgsfield, Freepik, Tripo, Genmo); the card's own estimate for this part is US$2,140M-3,590M-5,660M (low-central-high). Low and high shares hold the denominator at its central."},"avatar-video":{"participates":false,"evidence":{"url":"https://news.adobe.com/news/2026/06/adobe-to-acquire-topaz-labs","title":"Adobe to Acquire Topaz Labs","publisher":"Adobe","publishedDate":"2026-06-25","accessedDate":"2026-10-22","tier":"T2","locator":"Adobe's release lists Topaz Photo, Video, Gigapixel, Astra and Bloom (image and video enhancement); no speech, music, avatar or robot models."}},"robotics-science":{"participates":false,"evidence":{"url":"https://news.adobe.com/news/2026/06/adobe-to-acquire-topaz-labs","title":"Adobe to Acquire Topaz Labs","publisher":"Adobe","publishedDate":"2026-06-25","accessedDate":"2026-10-22","tier":"T2","locator":"Adobe's release lists Topaz Photo, Video, Gigapixel, Astra and Bloom (image and video enhancement); no speech, music, avatar or robot models."}}},"screening":{"image-video":{"levelLow":1,"levelHigh":2}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"9514e069c8c2c6fe3fc7313150580d5404d2f5780d96ed2c3c4dcd9cf6594d3d"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"9514e069c8c2c6fe3fc7313150580d5404d2f5780d96ed2c3c4dcd9cf6594d3d"}}},"assemblyai":{"schema":"5lc-index-inputs/1","key":"models-specialist-models","entityId":"assemblyai","name":"AssemblyAI","kind":"private","commercialActivityVerified":true,"shares":{"speech-voice":{"low":4.0454,"central":4.0454,"high":4.0454,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Annualized revenue, October 2026 (month in progress)","periodEnd":"2026-10-22","source":{"url":"https://sacra.com/c/assemblyai/","title":"AssemblyAI (Sacra company profile)","publisher":"Sacra","publishedDate":"undated","accessedDate":"2026-10-22","tier":"T4","locator":"Sacra estimates annualized revenue of $26M (end-2025), $32M (February 2026) and $76M (October 2026)"},"note":"Market-research estimate by Sacra: $76M annualized revenue in October 2026 (month only; $26M end-2025, $32M February 2026). Sacra's profile describes speech-to-text APIs and a voice agent API. Share = the central US$ figure ÷ the part total of US$1,879M, which is the register candidates' central figures (US$1,779M) plus an outside-register allowance of US$100M for other speech developers not in the register (for example Gladia, Soniox, Fish Audio, Murf, WellSaid, Rev); the card's own estimate for this part is US$1,050M-1,440M-2,280M (low-central-high). Low and high shares hold the denominator at its central."},"music":{"participates":false,"evidence":{"url":"https://sacra.com/c/assemblyai/","title":"AssemblyAI (Sacra company profile)","publisher":"Sacra","publishedDate":"undated","accessedDate":"2026-10-22","tier":"T4","locator":"Sacra describes AssemblyAI as selling speech-to-text, speech understanding and voice agent APIs plus an LLM gateway; no music, image, video, avatar or robot models."}},"image-video":{"participates":false,"evidence":{"url":"https://sacra.com/c/assemblyai/","title":"AssemblyAI (Sacra company profile)","publisher":"Sacra","publishedDate":"undated","accessedDate":"2026-10-22","tier":"T4","locator":"Sacra describes AssemblyAI as selling speech-to-text, speech understanding and voice agent APIs plus an LLM gateway; no music, image, video, avatar or robot models."}},"avatar-video":{"participates":false,"evidence":{"url":"https://sacra.com/c/assemblyai/","title":"AssemblyAI (Sacra company profile)","publisher":"Sacra","publishedDate":"undated","accessedDate":"2026-10-22","tier":"T4","locator":"Sacra describes AssemblyAI as selling speech-to-text, speech understanding and voice agent APIs plus an LLM gateway; no music, image, video, avatar or robot models."}},"robotics-science":{"participates":false,"evidence":{"url":"https://sacra.com/c/assemblyai/","title":"AssemblyAI (Sacra company profile)","publisher":"Sacra","publishedDate":"undated","accessedDate":"2026-10-22","tier":"T4","locator":"Sacra describes AssemblyAI as selling speech-to-text, speech understanding and voice agent APIs plus an LLM gateway; no music, image, video, avatar or robot models."}}},"priorShares":{"speech-voice":{"low":3.0995,"central":3.0995,"high":3.0995,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Annualised revenue, end of 2025 (Sacra estimate)","periodEnd":"2025-12-31","source":{"url":"https://sacra.com/c/assemblyai/","title":"AssemblyAI (Sacra company profile)","publisher":"Sacra","publishedDate":"undated","accessedDate":"2026-10-22","tier":"T4","locator":"Sacra estimates AssemblyAI reached US$26M of annualised revenue at the end of 2025 (also US$8.6M at end-2024 and US$32M in February 2026)"},"note":"Sacra's own estimate of annualised revenue (T4); the page states no method and dates no price list or customer count. The current US$76M (October 2026) is from the same page and series, so the two compare. Prior share = the prior US$ figure ÷ the prior part total of US$839M, which is the current part total of US$1,879M ÷ 2.24, the growth of this part over the year: pooled from the four speech companies with a sourced current and prior figure (ElevenLabs US$600M ÷ US$200M, AssemblyAI US$76M ÷ US$26M, SoundHound's counted share US$117.7M ÷ US$94.2M, iFLYTEK's counted share US$83.5M ÷ US$70.8M), US$877.2M ÷ US$391.0M = 2.24. Companies without a sourced prior are assumed to have grown at that rate. Low and high hold the denominator at its central."}},"activityGrowth":{"current":{"low":76000000,"central":76000000,"high":76000000,"unit":"USD","basis":"Annualised revenue of AssemblyAI, US dollars (Sacra estimate)","period":"Annualised revenue, October 2026 (Sacra estimate)","periodEnd":"2026-10-22","source":{"url":"https://sacra.com/c/assemblyai/","title":"AssemblyAI (Sacra company profile)","publisher":"Sacra","publishedDate":"undated","accessedDate":"2026-10-22","tier":"T4","locator":"Sacra estimates AssemblyAI hit US$76M in annualised revenue in October 2026"},"note":"Sacra estimate (T4), no method stated; the figure is for a month still in progress at the access date."},"prior":{"low":26000000,"central":26000000,"high":26000000,"unit":"USD","basis":"Annualised revenue of AssemblyAI, US dollars (Sacra estimate)","period":"Annualised revenue, end of 2025 (Sacra estimate)","periodEnd":"2025-12-31","source":{"url":"https://sacra.com/c/assemblyai/","title":"AssemblyAI (Sacra company profile)","publisher":"Sacra","publishedDate":"undated","accessedDate":"2026-10-22","tier":"T4","locator":"Sacra estimates US$26M of annualised revenue at the end of 2025"},"note":"Same Sacra series as the current figure. The gap is 9 months 22 days, inside the 9 to 15 month window, so the growth of 76 ÷ 26 = 2.9 times covers a little under a year."}},"baselineRevenueUsd":{"low":76000000,"central":76000000,"high":76000000,"unit":"USD","basis":"Annualised revenue of AssemblyAI, US dollars (Sacra estimate)","period":"Annualised revenue, October 2026 (Sacra estimate)","periodEnd":"2026-10-22","source":{"url":"https://sacra.com/c/assemblyai/","title":"AssemblyAI (Sacra company profile)","publisher":"Sacra","publishedDate":"undated","accessedDate":"2026-10-22","tier":"T4","locator":"Sacra estimates AssemblyAI hit US$76M in annualised revenue in October 2026"},"note":"Sacra estimate (T4)."},"screening":{"speech-voice":{"levelLow":1,"levelHigh":2}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"7b23590b28447016385793b9e5cccf348c3aace1e66318efd915875ce6d57ed2"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"7b23590b28447016385793b9e5cccf348c3aace1e66318efd915875ce6d57ed2"}}},"xai":{"schema":"5lc-index-inputs/1","key":"models-specialist-models","entityId":"xai","name":"xAI (Grok Imagine)","kind":"unit","parentId":"spacex","commercialActivityVerified":true,"shares":{"speech-voice":{"low":0,"central":0.0532,"high":0.5322,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Estimate for October 2026","periodEnd":"2026-10-22","source":{"url":"https://docs.x.ai/docs/models","title":"Models and pricing (xAI API documentation)","publisher":"xAI","publishedDate":"undated","tier":"T5","locator":"Text to speech US$15 per million characters, speech to text US$0.10-0.20 an hour, grok-voice-3.0 speech to speech US$0.08 a minute","accessedDate":"2026-10-22"},"note":"5LayerCake estimate (T5). xAI's API price list (opened 22 October 2026) sells text to speech, speech to text and speech to speech; no revenue is disclosed. Allowance for undisclosed revenue, not a measured figure: central US$1M, high US$10M. Share = the central US$ figure ÷ the part total of US$1,879M. That is the screening total (the register candidates' central figures of US$1,779M plus an outside-register allowance of US$100M for other developers not in the register), held fixed so every company in this sublayer is measured against one total; the card's own estimate for this part is US$1,050M-1,440M-2,280M (low-central-high). Low and high shares hold the denominator at its central."},"music":{"participates":false,"evidence":{"url":"https://docs.x.ai/docs/models","title":"Models and pricing (xAI API documentation)","publisher":"xAI","publishedDate":"undated","tier":"T2","locator":"The model list shows language, image (grok-imagine-image), video (grok-imagine-video), voice (grok-voice-3.0), text-to-speech and speech-to-text models; no music, avatar-video, robot-control or science model","accessedDate":"2026-10-22"}},"image-video":{"low":0.2661,"central":1.5965,"high":6.6519,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Estimate for October 2026","periodEnd":"2026-09-17","source":{"url":"https://vercel.com/blog/ai-gateway-production-index-september-2026","title":"Open-weight models take 56% of token volume, Astra doubles Fable 5.1 spend (AI Gateway Production Index, September 2026)","publisher":"Vercel","publishedDate":"2026-09-17","accessedDate":"2026-10-22","tier":"T5","locator":"August 2026 data: Nano Banana 50% of image spend vs GPT Image 44%; Seedance first in videos and video spend; Veo 20% of video spend; Grok Imagine 19% of videos"},"note":"5LayerCake estimate (T5). xAI (now inside SpaceX) opened the Grok Imagine API in January 2026; no revenue is disclosed. Vercel's gateway data shows Grok Imagine at 42% of videos generated in June, 31% in July and 19% in August 2026 (counts, not spend, and the model is cheap per second). Allowance of about 1.5% of the images-and-video part. Unit figure, not SpaceX's total. Share = the central US$ figure ÷ the part total of US$3,758M, which is the register candidates' central figures (US$3,608M) plus an outside-register allowance of US$150M for other image and video developers not in the register (for example Higgsfield, Freepik, Tripo, Genmo); the card's own estimate for this part is US$2,140M-3,590M-5,660M (low-central-high). Low and high shares hold the denominator at its central."},"avatar-video":{"participates":false,"evidence":{"url":"https://docs.x.ai/docs/models","title":"Models and pricing (xAI API documentation)","publisher":"xAI","publishedDate":"undated","tier":"T2","locator":"The model list shows language, image (grok-imagine-image), video (grok-imagine-video), voice (grok-voice-3.0), text-to-speech and speech-to-text models; no music, avatar-video, robot-control or science model","accessedDate":"2026-10-22"}},"robotics-science":{"participates":false,"evidence":{"url":"https://docs.x.ai/docs/models","title":"Models and pricing (xAI API documentation)","publisher":"xAI","publishedDate":"undated","tier":"T2","locator":"The model list shows language, image (grok-imagine-image), video (grok-imagine-video), voice (grok-voice-3.0), text-to-speech and speech-to-text models; no music, avatar-video, robot-control or science model","accessedDate":"2026-10-22"}}},"screening":{"speech-voice":{"levelLow":1,"levelHigh":2},"image-video":{"levelLow":1,"levelHigh":2}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"b4dd0dcc4eabcaa94e46f3429b79d019f4588a5e1f7c32ecfa1d66f22a5cb3e0"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"b4dd0dcc4eabcaa94e46f3429b79d019f4588a5e1f7c32ecfa1d66f22a5cb3e0"}}},"meshy":{"schema":"5lc-index-inputs/1","key":"models-specialist-models","entityId":"meshy","name":"Meshy","kind":"private","commercialActivityVerified":true,"shares":{"speech-voice":{"participates":false,"evidence":{"url":"https://www.prnewswire.com/news-releases/meshy-surpasses-100-million-in-annual-recurring-revenue-growing-100x-in-under-two-years-302894078.html","title":"Meshy Surpasses $100 Million in Annual Recurring Revenue, Growing 100x in Under Two Years","publisher":"Meshy (PR Newswire)","publishedDate":"2026-09-30","tier":"T2","locator":"The release describes Meshy as turning text and images into production-ready 3D assets (text-to-3D, image-to-3D, a phone app); it names no speech, music, avatar-video or robot-model product","accessedDate":"2026-10-22"}},"music":{"participates":false,"evidence":{"url":"https://www.prnewswire.com/news-releases/meshy-surpasses-100-million-in-annual-recurring-revenue-growing-100x-in-under-two-years-302894078.html","title":"Meshy Surpasses $100 Million in Annual Recurring Revenue, Growing 100x in Under Two Years","publisher":"Meshy (PR Newswire)","publishedDate":"2026-09-30","tier":"T2","locator":"The release describes Meshy as turning text and images into production-ready 3D assets (text-to-3D, image-to-3D, a phone app); it names no speech, music, avatar-video or robot-model product","accessedDate":"2026-10-22"}},"image-video":{"low":0,"central":1.3305,"high":2.661,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"ARR, September 2026, share counted","periodEnd":"2026-09-30","source":{"url":"https://www.prnewswire.com/news-releases/meshy-surpasses-100-million-in-annual-recurring-revenue-growing-100x-in-under-two-years-302894078.html","title":"Meshy Surpasses $100 Million in Annual Recurring Revenue, Growing 100x in Under Two Years","publisher":"Meshy (PR Newswire)","publishedDate":"2026-09-30","tier":"T5","locator":"\"As of September 2026, Meshy's annual recurring revenue (ARR) has surpassed $100 million\" (dateline 30 September 2026)","accessedDate":"2026-10-22"},"note":"5LayerCake estimate (T5) from a company figure (T2). Meshy says its ARR has surpassed US$100M as of September 2026 (release of 30 September 2026). It sells text-to-3D and image-to-3D models for games, film and 3D printing. The card names images and video, not 3D, so the counted share is a card call: low 0%, central 50% (US$50M), high 100% (US$100M). 'Surpassed' makes US$100M a floor. Share = the central US$ figure ÷ the part total of US$3,758M. That is the screening total (the register candidates' central figures of US$3,608M plus an outside-register allowance of US$150M for other developers not in the register), held fixed so every company in this sublayer is measured against one total; the card's own estimate for this part is US$2,140M-3,590M-5,660M (low-central-high). Low and high shares hold the denominator at its central."},"avatar-video":{"participates":false,"evidence":{"url":"https://www.prnewswire.com/news-releases/meshy-surpasses-100-million-in-annual-recurring-revenue-growing-100x-in-under-two-years-302894078.html","title":"Meshy Surpasses $100 Million in Annual Recurring Revenue, Growing 100x in Under Two Years","publisher":"Meshy (PR Newswire)","publishedDate":"2026-09-30","tier":"T2","locator":"The release describes Meshy as turning text and images into production-ready 3D assets (text-to-3D, image-to-3D, a phone app); it names no speech, music, avatar-video or robot-model product","accessedDate":"2026-10-22"}},"robotics-science":{"participates":false,"evidence":{"url":"https://www.prnewswire.com/news-releases/meshy-surpasses-100-million-in-annual-recurring-revenue-growing-100x-in-under-two-years-302894078.html","title":"Meshy Surpasses $100 Million in Annual Recurring Revenue, Growing 100x in Under Two Years","publisher":"Meshy (PR Newswire)","publishedDate":"2026-09-30","tier":"T2","locator":"The release describes Meshy as turning text and images into production-ready 3D assets (text-to-3D, image-to-3D, a phone app); it names no speech, music, avatar-video or robot-model product","accessedDate":"2026-10-22"}}},"priorShares":{"image-video":{"low":0,"central":0.485,"high":0.9699,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"ARR, November 2025, share counted","periodEnd":"2025-11-14","source":{"url":"https://www.prnewswire.com/de/pressemitteilungen/meshy-erzielt-einen-jahrlichen-wiederkehrenden-umsatz-arr-von-15-mio-usd-mit-einem-monatlichen-wachstum-von-30--und-prasentiert-meshy-6-preview-fur-die-3d-erstellung-der-nachsten-generation-302615821.html","title":"Meshy erzielt einen jährlichen wiederkehrenden Umsatz (ARR) von 15 Mio. USD mit einem monatlichen Wachstum von 30 % und präsentiert Meshy 6 Preview für die 3D-Erstellung der nächsten Generation","publisher":"Meshy (PR Newswire)","publishedDate":"2025-11-14","tier":"T5","locator":"\"In weniger als zwei Jahren seit dem Start hat Meshy einen ARR von 15 Mio. USD erreicht\" (dateline 14 November 2025)","accessedDate":"2026-10-22"},"note":"5LayerCake estimate (T5) from a company figure (T2). Meshy said in its release of 14 November 2025 that it had reached US$15M of ARR; the same 0% / 50% / 100% counted share as the current figure gives low 0, central US$7.5M, high US$15M. A March 2026 release (9 March 2026, https://www.prnewswire.com/news-releases/meshy-unveils-meshy-labs-at-gdc-2026--breakthrough-ai-native-gameplay-and-30m-arr-milestone-302708046.html) put ARR at US$30M, consistent with this path. Prior share = the prior US$ figure ÷ the prior part total of US$1,547M, which is the current part total of US$3,758M ÷ 2.43, the growth of this part over the year: pooled from every company in this sublayer's file set with a sourced pair, because the image-and-video pairs alone (Kling US$500M ÷ US$138M, Meshy at 50% US$50M ÷ US$7.5M, MiniMax's video share US$30M ÷ US$15M) cover only 15% of this part: the four speech pairs (ElevenLabs US$600M ÷ US$200M, AssemblyAI US$76M ÷ US$26M, SoundHound US$117.7M ÷ US$94.2M, iFLYTEK US$83.5M ÷ US$70.8M) and the two avatar pairs (HeyGen US$205M ÷ US$95M, Synthesia US$150M ÷ US$100M) plus these, US$1,812M ÷ US$747M = 2.43. Companies without a sourced prior are assumed to have grown at that rate. Low and high hold the denominator at its central."}},"activityGrowth":{"current":{"low":100000000,"central":100000000,"high":100000000,"unit":"USD","basis":"Annual recurring revenue (ARR) of Meshy, US dollars (company statements)","period":"ARR, September 2026","periodEnd":"2026-09-30","source":{"url":"https://www.prnewswire.com/news-releases/meshy-surpasses-100-million-in-annual-recurring-revenue-growing-100x-in-under-two-years-302894078.html","title":"Meshy Surpasses $100 Million in Annual Recurring Revenue, Growing 100x in Under Two Years","publisher":"Meshy (PR Newswire)","publishedDate":"2026-09-30","tier":"T2","locator":"\"As of September 2026, Meshy's annual recurring revenue (ARR) has surpassed $100 million\" (dateline 30 September 2026)","accessedDate":"2026-10-22"},"note":"The company says ARR has surpassed US$100M. A floor. All of Meshy's ARR, whatever share counts as images and video."},"prior":{"low":15000000,"central":15000000,"high":15000000,"unit":"USD","basis":"Annual recurring revenue (ARR) of Meshy, US dollars (company statements)","period":"ARR, November 2025","periodEnd":"2025-11-14","source":{"url":"https://www.prnewswire.com/de/pressemitteilungen/meshy-erzielt-einen-jahrlichen-wiederkehrenden-umsatz-arr-von-15-mio-usd-mit-einem-monatlichen-wachstum-von-30--und-prasentiert-meshy-6-preview-fur-die-3d-erstellung-der-nachsten-generation-302615821.html","title":"Meshy erzielt einen jährlichen wiederkehrenden Umsatz (ARR) von 15 Mio. USD mit einem monatlichen Wachstum von 30 % und präsentiert Meshy 6 Preview für die 3D-Erstellung der nächsten Generation","publisher":"Meshy (PR Newswire)","publishedDate":"2025-11-14","tier":"T2","locator":"\"In weniger als zwei Jahren seit dem Start hat Meshy einen ARR von 15 Mio. USD erreicht\" (dateline 14 November 2025)","accessedDate":"2026-10-22"},"note":"The company says it had reached US$15M of ARR (release of 14 November 2025, in German). 100 ÷ 15 = 6.7 times; the engine clamps growth at +300%."}},"baselineRevenueUsd":{"low":100000000,"central":100000000,"high":100000000,"unit":"USD","basis":"Annual recurring revenue (ARR) of Meshy, US dollars (company statements)","period":"ARR, September 2026","periodEnd":"2026-09-30","source":{"url":"https://www.prnewswire.com/news-releases/meshy-surpasses-100-million-in-annual-recurring-revenue-growing-100x-in-under-two-years-302894078.html","title":"Meshy Surpasses $100 Million in Annual Recurring Revenue, Growing 100x in Under Two Years","publisher":"Meshy (PR Newswire)","publishedDate":"2026-09-30","tier":"T2","locator":"\"As of September 2026, Meshy's annual recurring revenue (ARR) has surpassed $100 million\" (dateline 30 September 2026)","accessedDate":"2026-10-22"},"note":"The company says ARR has surpassed US$100M (3D assets; only part counts as images and video for the share)."},"screening":{"image-video":{"levelLow":1,"levelHigh":2}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"4cb58435b2ccea220419c3800d2d5626c22a49db6642838baea37f7be9971b6c"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"4cb58435b2ccea220419c3800d2d5626c22a49db6642838baea37f7be9971b6c"}}},"photoroom":{"schema":"5lc-index-inputs/1","key":"models-specialist-models","entityId":"photoroom","name":"Photoroom","kind":"private","commercialActivityVerified":true,"shares":{"speech-voice":{"participates":false,"evidence":{"url":"https://sacra.com/c/photoroom/","title":"Photoroom (Sacra company profile)","publisher":"Sacra","publishedDate":"undated","accessedDate":"2026-10-22","tier":"T4","locator":"Sacra describes Photoroom as an image-editing app and API; no speech, music, avatar or robot models."}},"music":{"participates":false,"evidence":{"url":"https://sacra.com/c/photoroom/","title":"Photoroom (Sacra company profile)","publisher":"Sacra","publishedDate":"undated","accessedDate":"2026-10-22","tier":"T4","locator":"Sacra describes Photoroom as an image-editing app and API; no speech, music, avatar or robot models."}},"image-video":{"low":0.7982,"central":1.49,"high":2.6608,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"ARR 2024 (Sacra), own-model share","periodEnd":"2024-12-31","source":{"url":"https://sacra.com/c/photoroom/","title":"Photoroom (Sacra company profile)","publisher":"Sacra","publishedDate":"undated","accessedDate":"2026-10-22","tier":"T5","locator":"Sacra estimates about $94M ARR for 2024 (up 89%); freemium subscriptions plus API; no 2025 or 2026 estimate"},"note":"5LayerCake estimate (T5). Sacra estimates about US$94M ARR for 2024 (up 89%), from a freemium app plus API; no 2025 or 2026 figure. Only revenue from Photoroom's own image models counts, taken as 30-100% of ARR (central 60%): the boundary with the Applications part is a card call. The date is more than 12 months old, so the engine widens it. Share = the central US$ figure ÷ the part total of US$3,758M, which is the register candidates' central figures (US$3,608M) plus an outside-register allowance of US$150M for other image and video developers not in the register (for example Higgsfield, Freepik, Tripo, Genmo); the card's own estimate for this part is US$2,140M-3,590M-5,660M (low-central-high). Low and high shares hold the denominator at its central."},"avatar-video":{"participates":false,"evidence":{"url":"https://sacra.com/c/photoroom/","title":"Photoroom (Sacra company profile)","publisher":"Sacra","publishedDate":"undated","accessedDate":"2026-10-22","tier":"T4","locator":"Sacra describes Photoroom as an image-editing app and API; no speech, music, avatar or robot models."}},"robotics-science":{"participates":false,"evidence":{"url":"https://sacra.com/c/photoroom/","title":"Photoroom (Sacra company profile)","publisher":"Sacra","publishedDate":"undated","accessedDate":"2026-10-22","tier":"T4","locator":"Sacra describes Photoroom as an image-editing app and API; no speech, music, avatar or robot models."}}},"baselineRevenueUsd":{"low":94000000,"central":94000000,"high":94000000,"unit":"USD","basis":"Annual recurring revenue (ARR) of Photoroom, US dollars (Sacra estimate)","period":"ARR, 2024 (Sacra estimate)","periodEnd":"2024-12-31","source":{"url":"https://sacra.com/c/photoroom/","title":"Photoroom (Sacra company profile)","publisher":"Sacra","publishedDate":"undated","accessedDate":"2026-10-22","tier":"T4","locator":"Sacra estimates about US$94M of ARR for 2024, up 89%"},"note":"Sacra estimate (T4); more than 12 months old, so the engine widens it. Includes the app's non-model revenue; the share used for scoring counts only 30-100% of it."},"screening":{"image-video":{"levelLow":1,"levelHigh":2}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"74cd98f56fd4070983ccaeab97f879d65bb1aebdfd813ac8cfb7ae2d8190c933"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"74cd98f56fd4070983ccaeab97f879d65bb1aebdfd813ac8cfb7ae2d8190c933"}}},"pixverse":{"schema":"5lc-index-inputs/1","key":"models-specialist-models","entityId":"pixverse","name":"AIsphere (PixVerse)","kind":"private","commercialActivityVerified":true,"shares":{"speech-voice":{"participates":false,"evidence":{"url":"https://www.kr-asia.com/aisphere-touts-pixverse-as-the-canva-for-video-generation-and-lands-the-funding-to-prove-it","title":"AIsphere touts PixVerse as the \"Canva for video generation,\" and lands the funding to prove it","publisher":"KrASIA (from 36Kr)","publishedDate":"2025-09-17","accessedDate":"2026-10-22","tier":"T3","locator":"KrASIA describes PixVerse as a video generation product; no speech, music, avatar or robot models."}},"music":{"participates":false,"evidence":{"url":"https://www.kr-asia.com/aisphere-touts-pixverse-as-the-canva-for-video-generation-and-lands-the-funding-to-prove-it","title":"AIsphere touts PixVerse as the \"Canva for video generation,\" and lands the funding to prove it","publisher":"KrASIA (from 36Kr)","publishedDate":"2025-09-17","accessedDate":"2026-10-22","tier":"T3","locator":"KrASIA describes PixVerse as a video generation product; no speech, music, avatar or robot models."}},"image-video":{"low":1.0644,"central":1.3305,"high":1.9957,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"ARR, 2026 (reported July 2026), floor and growth","periodEnd":"2026-07-14","source":{"url":"https://startupfortune.com/pixverse-tops-a-2-billion-valuation-after-extending-its-series-c-to-439-million/","title":"PixVerse tops a $2 billion valuation after extending its Series C to $439 million","publisher":"Startup Fortune","publishedDate":"2026-07-14","tier":"T5","locator":"\"Annual recurring revenue sits above $40 million\"; \"Bloomberg reported that figure is growing more than tenfold year over year\"","accessedDate":"2026-10-22"},"note":"5LayerCake estimate (T5) from a press report (T3). Startup Fortune (14 July 2026) says PixVerse's annual recurring revenue 'sits above $40 million' and that Bloomberg reported the figure growing more than tenfold year over year; the article does not date the figure or say whether it comes from the company, and TechCrunch's report of the same round (13 July 2026) gives no revenue figure. Low = the US$40M floor; central US$50M allows a quarter of growth on the floor; high US$75M allows growth at the reported tenfold-a-year pace (about 21% a month) for the same quarter. Share = the central US$ figure ÷ the part total of US$3,758M. That is the screening total (the register candidates' central figures of US$3,608M plus an outside-register allowance of US$150M for other developers not in the register), held fixed so every company in this sublayer is measured against one total; the card's own estimate for this part is US$2,140M-3,590M-5,660M (low-central-high). Low and high shares hold the denominator at its central."},"avatar-video":{"participates":false,"evidence":{"url":"https://www.kr-asia.com/aisphere-touts-pixverse-as-the-canva-for-video-generation-and-lands-the-funding-to-prove-it","title":"AIsphere touts PixVerse as the \"Canva for video generation,\" and lands the funding to prove it","publisher":"KrASIA (from 36Kr)","publishedDate":"2025-09-17","accessedDate":"2026-10-22","tier":"T3","locator":"KrASIA describes PixVerse as a video generation product; no speech, music, avatar or robot models."}},"robotics-science":{"participates":false,"evidence":{"url":"https://www.kr-asia.com/aisphere-touts-pixverse-as-the-canva-for-video-generation-and-lands-the-funding-to-prove-it","title":"AIsphere touts PixVerse as the \"Canva for video generation,\" and lands the funding to prove it","publisher":"KrASIA (from 36Kr)","publishedDate":"2025-09-17","accessedDate":"2026-10-22","tier":"T3","locator":"KrASIA describes PixVerse as a video generation product; no speech, music, avatar or robot models."}}},"priorShares":{"image-video":{"low":0.1293,"central":0.2586,"high":0.2586,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"ARR, July 2025 (implied by 'more than tenfold')","periodEnd":"2025-07-14","source":{"url":"https://startupfortune.com/pixverse-tops-a-2-billion-valuation-after-extending-its-series-c-to-439-million/","title":"PixVerse tops a $2 billion valuation after extending its Series C to $439 million","publisher":"Startup Fortune","publishedDate":"2026-07-14","tier":"T5","locator":"\"Annual recurring revenue sits above $40 million\"; \"Bloomberg reported that figure is growing more than tenfold year over year\"","accessedDate":"2026-10-22"},"note":"5LayerCake derivation (T5). Startup Fortune reports ARR above US$40M, growing more than tenfold in a year (Bloomberg): so the figure a year earlier was below 40 ÷ 10 = US$4M. Central and high use that ceiling (US$4M), low US$2M. Because the engine clamps growth at +300%, any prior below US$10M gives the same growth score. Prior share = the prior US$ figure ÷ the prior part total of US$1,547M, which is the current part total of US$3,758M ÷ 2.43, the growth of this part over the year: pooled from every company in this sublayer's file set with a sourced pair, because the image-and-video pairs alone (Kling US$500M ÷ US$138M, Meshy at 50% US$50M ÷ US$7.5M, MiniMax's video share US$30M ÷ US$15M) cover only 15% of this part: the four speech pairs (ElevenLabs US$600M ÷ US$200M, AssemblyAI US$76M ÷ US$26M, SoundHound US$117.7M ÷ US$94.2M, iFLYTEK US$83.5M ÷ US$70.8M) and the two avatar pairs (HeyGen US$205M ÷ US$95M, Synthesia US$150M ÷ US$100M) plus these, US$1,812M ÷ US$747M = 2.43. Companies without a sourced prior are assumed to have grown at that rate. Low and high hold the denominator at its central."}},"activityGrowth":{"current":{"low":40000000,"central":40000000,"high":40000000,"unit":"USD","basis":"Annual recurring revenue (ARR) of PixVerse, US dollars (press report of company figures)","period":"ARR, 2026 (reported July 2026)","periodEnd":"2026-07-14","source":{"url":"https://startupfortune.com/pixverse-tops-a-2-billion-valuation-after-extending-its-series-c-to-439-million/","title":"PixVerse tops a $2 billion valuation after extending its Series C to $439 million","publisher":"Startup Fortune","publishedDate":"2026-07-14","tier":"T3","locator":"\"Annual recurring revenue sits above $40 million\"; \"Bloomberg reported that figure is growing more than tenfold year over year\"","accessedDate":"2026-10-22"},"note":"'Above $40 million', so a floor; the article does not date the figure."},"prior":{"low":4000000,"central":4000000,"high":4000000,"unit":"USD","basis":"Annual recurring revenue (ARR) of PixVerse, US dollars (press report of company figures)","period":"ARR, July 2025 (implied)","periodEnd":"2025-07-14","source":{"url":"https://startupfortune.com/pixverse-tops-a-2-billion-valuation-after-extending-its-series-c-to-439-million/","title":"PixVerse tops a $2 billion valuation after extending its Series C to $439 million","publisher":"Startup Fortune","publishedDate":"2026-07-14","tier":"T5","locator":"'Bloomberg reported that figure is growing more than tenfold year over year'","accessedDate":"2026-10-22"},"note":"Derived (T5): 40 ÷ 10 = 4, a ceiling for the earlier figure because the report says 'more than tenfold'."}},"baselineRevenueUsd":{"low":40000000,"central":40000000,"high":40000000,"unit":"USD","basis":"Annual recurring revenue (ARR) of PixVerse, US dollars (press report of company figures)","period":"ARR, 2026 (reported July 2026)","periodEnd":"2026-07-14","source":{"url":"https://startupfortune.com/pixverse-tops-a-2-billion-valuation-after-extending-its-series-c-to-439-million/","title":"PixVerse tops a $2 billion valuation after extending its Series C to $439 million","publisher":"Startup Fortune","publishedDate":"2026-07-14","tier":"T3","locator":"\"Annual recurring revenue sits above $40 million\"; \"Bloomberg reported that figure is growing more than tenfold year over year\"","accessedDate":"2026-10-22"},"note":"'Above $40 million': a floor."},"screening":{"image-video":{"levelLow":1,"levelHigh":2}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"708a93e098857c49934b6f2ee9894d7ebd1b2f0dabbeab3520c4355eefef52f0"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"708a93e098857c49934b6f2ee9894d7ebd1b2f0dabbeab3520c4355eefef52f0"}}},"alibaba-qwen":{"schema":"5lc-index-inputs/1","key":"models-specialist-models","entityId":"alibaba-qwen","name":"Alibaba Qwen (Wan, Qwen-Image, speech)","kind":"unit","parentId":"alibaba","commercialActivityVerified":false,"shares":{"speech-voice":{"low":0.1064,"central":0.5322,"high":2.1288,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Estimate for October 2026","periodEnd":"2026-10-22","source":{"url":"https://www.alibabacloud.com/help/en/model-studio/model-pricing","title":"Model pricing (Alibaba Cloud Model Studio documentation)","publisher":"Alibaba Cloud","publishedDate":"undated","tier":"T5","locator":"Qwen-TTS and CosyVoice text to speech (about US$0.10-0.30 per 10,000 characters), Qwen / Fun-ASR speech recognition (about US$0.000012-0.00009 a second) and voice cloning (US$0.01 a clone), billed pay-as-you-go","accessedDate":"2026-10-22"},"note":"5LayerCake estimate (T5). Alibaba Cloud Model Studio's price list (opened 22 October 2026) sells Qwen3-TTS and CosyVoice speech synthesis, Qwen3-ASR, Fun-ASR and Paraformer recognition and voice cloning, billed pay-as-you-go; it gives no usage or revenue. Allowance only: central US$10M, high US$40M. Unit figure, not Alibaba's total. Share = the central US$ figure ÷ the part total of US$1,879M. That is the screening total (the register candidates' central figures of US$1,779M plus an outside-register allowance of US$100M for other developers not in the register), held fixed so every company in this sublayer is measured against one total; the card's own estimate for this part is US$1,050M-1,440M-2,280M (low-central-high). Low and high shares hold the denominator at its central."},"music":{"low":0,"central":0.1302,"high":1.3021,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Estimate for October 2026","periodEnd":"2026-10-22","source":{"url":"https://www.alibabacloud.com/help/en/model-studio/model-pricing","title":"Model pricing (Alibaba Cloud Model Studio documentation)","publisher":"Alibaba Cloud","publishedDate":"undated","tier":"T5","locator":"fun-music-v1 at US$0.000275 and fun-music-preview at US$0.000695 per second of output audio (Beijing)","accessedDate":"2026-10-22"},"note":"5LayerCake estimate (T5). The Model Studio price list (opened 22 October 2026) includes Fun-Music generation models priced per second of output audio; no usage or revenue is given. Allowance for undisclosed revenue, not a measured figure: central US$0.5M, high US$5M. Unit figure, not Alibaba's total. Share = the central US$ figure ÷ the part total of US$384M. That is the screening total (the register candidates' central figures of US$364M plus an outside-register allowance of US$20M for other developers not in the register), held fixed so every company in this sublayer is measured against one total; the card's own estimate for this part is US$315M-336M-500M (low-central-high). Low and high shares hold the denominator at its central."},"image-video":{"low":0.2129,"central":1.0644,"high":3.9915,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Estimate for October 2026","periodEnd":"2026-10-22","source":{"url":"https://www.alibabacloud.com/help/en/model-studio/model-pricing","title":"Model pricing (Alibaba Cloud Model Studio documentation)","publisher":"Alibaba Cloud","publishedDate":"undated","tier":"T5","locator":"Wan video models (for example wan3.0-video-prime, US$0.068-0.28 a second of output by resolution) and Wan / Qwen-Image image models (US$0.014-0.075 an image), billed pay-as-you-go","accessedDate":"2026-10-22"},"note":"5LayerCake estimate (T5). The Model Studio price list (opened 22 October 2026) sells the Wan video and image models and Qwen-Image pay-as-you-go; open Wan weights also earn through hosting by others (counted in Inference serving). No usage or revenue is given. Allowance of about 1% of the part: central US$40M, high US$150M. Unit figure, not Alibaba's total. Share = the central US$ figure ÷ the part total of US$3,758M. That is the screening total (the register candidates' central figures of US$3,608M plus an outside-register allowance of US$150M for other developers not in the register), held fixed so every company in this sublayer is measured against one total; the card's own estimate for this part is US$2,140M-3,590M-5,660M (low-central-high). Low and high shares hold the denominator at its central."},"avatar-video":{"low":0,"central":0.1006,"high":1.006,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Estimate for October 2026","periodEnd":"2026-10-22","source":{"url":"https://www.alibabacloud.com/help/en/model-studio/model-pricing","title":"Model pricing (Alibaba Cloud Model Studio documentation)","publisher":"Alibaba Cloud","publishedDate":"undated","tier":"T5","locator":"wan2.2-s2v (speech-to-video digital human) at US$0.072-0.129 a second of output and wan2.2-s2v-detect at US$0.000574 an input image (Beijing)","accessedDate":"2026-10-22"},"note":"5LayerCake estimate (T5). The Model Studio price list (opened 22 October 2026) sells a speech-to-video digital-human model (wan2.2-s2v); no usage or revenue is given. Allowance for undisclosed revenue, not a measured figure: central US$0.5M, high US$5M. Unit figure, not Alibaba's total. Share = the central US$ figure ÷ the part total of US$497M. That is the screening total (the register candidates' central figures of US$467M plus an outside-register allowance of US$30M for other developers not in the register), held fixed so every company in this sublayer is measured against one total; the card's own estimate for this part is US$400M-450M-550M (low-central-high). Low and high shares hold the denominator at its central."},"robotics-science":{"low":0,"central":0.1282,"high":1.2821,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Estimate for October 2026","periodEnd":"2026-10-22","source":{"url":"https://www.alibabacloud.com/help/en/model-studio/model-pricing","title":"Model pricing (Alibaba Cloud Model Studio documentation)","publisher":"Alibaba Cloud","publishedDate":"undated","tier":"T5","locator":"The price list names speech, music, image, video and digital-human models; none of the sections read names a robot-control or science model","accessedDate":"2026-10-22"},"note":"5LayerCake estimate (T5). No Alibaba robot-model product was found on the price list. Allowance for undisclosed revenue, not a measured figure: central US$0.3M, high US$3M. Unit figure, not Alibaba's total. Share = the central US$ figure ÷ the part total of US$234M. That is the screening total (the register candidates' central figures of US$219M plus an outside-register allowance of US$15M for other developers not in the register), held fixed so every company in this sublayer is measured against one total; the card's own estimate for this part is US$140M-200M-350M (low-central-high). Low and high shares hold the denominator at its central."}},"screening":{"speech-voice":{"levelLow":1,"levelHigh":2},"music":{"levelLow":1,"levelHigh":2},"image-video":{"levelLow":1,"levelHigh":2},"avatar-video":{"levelLow":1,"levelHigh":2},"robotics-science":{"levelLow":1,"levelHigh":2}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"29fd0b13015a3881bc226d407c85f7e7c47597637ed8afbfd3701d3c93bd3fb5"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"29fd0b13015a3881bc226d407c85f7e7c47597637ed8afbfd3701d3c93bd3fb5"}}},"minimax":{"schema":"5lc-index-inputs/1","key":"models-specialist-models","entityId":"minimax","name":"MiniMax (Hailuo, Speech, Music)","kind":"public","commercialActivityVerified":true,"shares":{"speech-voice":{"low":0.2129,"central":0.7984,"high":2.3953,"unit":"percent","basis":"Half-year 2026 revenue × 2 in US dollars, not the card's latest-quarter × 4 run-rate","offBasis":true,"period":"H1 2026 Open Platform revenue x 2, share counted","periodEnd":"2026-06-30","source":{"url":"https://www1.hkexnews.hk/listedco/listconews/sehk/2026/0826/2026082600680.pdf","title":"MiniMax Group Inc. interim results announcement for the six months ended June 30, 2026","publisher":"MiniMax Group (HKEXnews)","publishedDate":"2026-08-26","accessedDate":"2026-10-22","tier":"T5","locator":"Management discussion: H1 2026 revenue US$116.6m; Open Platform and enterprise services US$73.9m; AI-native products (Hailuo AI and others) US$42.6m; no split by modality"},"note":"5LayerCake estimate (T5) from a filing figure (T1). Open Platform and enterprise services revenue was US$73.9M in H1 2026 (US$148M a year), mostly language-model tokens; MiniMax Speech is sold by the character through the platform. Speech assumed 3-30% (central 10%). The results do not split speech out. Share = the central US$ figure ÷ the part total of US$1,879M, which is the register candidates' central figures (US$1,779M) plus an outside-register allowance of US$100M for other speech developers not in the register (for example Gladia, Soniox, Fish Audio, Murf, WellSaid, Rev); the card's own estimate for this part is US$1,050M-1,440M-2,280M (low-central-high). Low and high shares hold the denominator at its central."},"music":{"low":0.2602,"central":1.0409,"high":3.9032,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"H1 2026 estimate","periodEnd":"2026-06-30","source":{"url":"https://www1.hkexnews.hk/listedco/listconews/sehk/2026/0826/2026082600680.pdf","title":"MiniMax Group Inc. interim results announcement for the six months ended June 30, 2026","publisher":"MiniMax Group (HKEXnews)","publishedDate":"2026-08-26","accessedDate":"2026-10-22","tier":"T5","locator":"Management discussion: H1 2026 revenue US$116.6m; Open Platform and enterprise services US$73.9m; AI-native products (Hailuo AI and others) US$42.6m; no split by modality"},"note":"5LayerCake estimate (T5). MiniMax sells Music models through its platform (Audio & Music page); the results do not split music out. Allowance for undisclosed revenue, not a measured figure. Share = the central US$ figure ÷ the part total of US$384M, which is the register candidates' central figures (US$364M) plus an outside-register allowance of US$20M for other music-model developers not in the register (for example Producer.ai, AIVA, Soundraw); the card's own estimate for this part is US$315M-336M-500M (low-central-high). Low and high shares hold the denominator at its central."},"image-video":{"low":0.2661,"central":0.7982,"high":1.8625,"unit":"percent","basis":"Half-year 2026 revenue × 2 in US dollars, not the card's latest-quarter × 4 run-rate","offBasis":true,"period":"H1 2026 AI-native product revenue x 2, share counted","periodEnd":"2026-06-30","source":{"url":"https://www1.hkexnews.hk/listedco/listconews/sehk/2026/0826/2026082600680.pdf","title":"MiniMax Group Inc. interim results announcement for the six months ended June 30, 2026","publisher":"MiniMax Group (HKEXnews)","publishedDate":"2026-08-26","accessedDate":"2026-10-22","tier":"T5","locator":"Management discussion: H1 2026 revenue US$116.6m; Open Platform and enterprise services US$73.9m; AI-native products (Hailuo AI and others) US$42.6m; no split by modality"},"note":"5LayerCake estimate (T5) from a filing figure (T1). MiniMax's interim results give H1 2026 revenue of US$116.6M: AI-native products (Hailuo AI and others) US$42.6M and Open Platform US$73.9M, with no split by modality, and most platform growth is language-model tokens. AI-native products x 2 = US$85M a year; Hailuo video assumed 12-80% of it (central 35%) plus a little platform video. KrASIA (July 2025) had Hailuo above US$10M ARR. Talkie and agent products count in Applications. Further inputs: KrASIA (from 36Kr), 2025-07-11, https://kr-asia.com/chinas-video-ai-startups-are-doing-what-llm-firms-havent-making-money. Share = the central US$ figure ÷ the part total of US$3,758M, which is the register candidates' central figures (US$3,608M) plus an outside-register allowance of US$150M for other image and video developers not in the register (for example Higgsfield, Freepik, Tripo, Genmo); the card's own estimate for this part is US$2,140M-3,590M-5,660M (low-central-high). Low and high shares hold the denominator at its central."},"avatar-video":{"low":0,"central":0.1005,"high":1.0054,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"H1 2026 estimate","periodEnd":"2026-06-30","source":{"url":"https://www1.hkexnews.hk/listedco/listconews/sehk/2026/0826/2026082600680.pdf","title":"MiniMax Group Inc. interim results announcement for the six months ended June 30, 2026","publisher":"MiniMax Group (HKEXnews)","publishedDate":"2026-08-26","accessedDate":"2026-10-22","tier":"T5","locator":"Management discussion: H1 2026 revenue US$116.6m; Open Platform and enterprise services US$73.9m; AI-native products (Hailuo AI and others) US$42.6m; no split by modality"},"note":"5LayerCake estimate (T5). No MiniMax avatar-video product or revenue was found in the cited results. Allowance for undisclosed revenue, not a measured figure. Share = the central US$ figure ÷ the part total of US$497M, which is the register candidates' central figures (US$467M) plus an outside-register allowance of US$30M for other avatar-video developers not in the register (for example Akool, Argil, Creatify); the card's own estimate for this part is US$400M-450M-550M (low-central-high). Low and high shares hold the denominator at its central."},"robotics-science":{"low":0,"central":0.1284,"high":1.2837,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"H1 2026 estimate","periodEnd":"2026-06-30","source":{"url":"https://www1.hkexnews.hk/listedco/listconews/sehk/2026/0826/2026082600680.pdf","title":"MiniMax Group Inc. interim results announcement for the six months ended June 30, 2026","publisher":"MiniMax Group (HKEXnews)","publishedDate":"2026-08-26","accessedDate":"2026-10-22","tier":"T5","locator":"Management discussion: H1 2026 revenue US$116.6m; Open Platform and enterprise services US$73.9m; AI-native products (Hailuo AI and others) US$42.6m; no split by modality"},"note":"5LayerCake estimate (T5). No MiniMax robot-model product or revenue was found in the cited results. Allowance for undisclosed revenue, not a measured figure. Share = the central US$ figure ÷ the part total of US$234M, which is the register candidates' central figures (US$219M) plus an outside-register allowance of US$15M for other robot-model and science-model developers not in the register (for example Latent Labs, Periodic Labs, Absci); the card's own estimate for this part is US$140M-200M-350M (low-central-high). Low and high shares hold the denominator at its central."}},"priorShares":{"speech-voice":{"low":0.2129,"central":0.7984,"high":2.3953,"unit":"percent","basis":"Half-year 2025 revenue × 2 in US dollars, not the card's latest-quarter × 4 run-rate","offBasis":true,"period":"Held at the current share (prior period 2025-06-30)","periodEnd":"2025-06-30","source":{"url":"https://www1.hkexnews.hk/listedco/listconews/sehk/2026/0826/2026082600680.pdf","title":"MiniMax Group Inc. interim results announcement for the six months ended June 30, 2026","publisher":"MiniMax Group (HKEXnews)","publishedDate":"2026-08-26","accessedDate":"2026-10-22","tier":"T5","locator":"Management discussion: H1 2026 revenue US$116.6m; Open Platform and enterprise services US$73.9m; AI-native products (Hailuo AI and others) US$42.6m; no split by modality"},"note":"5LayerCake assumption (T5): no figure for the earlier period exists for this undisclosed allowance, so the share is held at the current share (the company is assumed to have grown with this part of the market). The H1 2025 Open Platform revenue (US$9.2M) is mostly language-model tokens, so its 703% growth is not used for speech."},"music":{"low":0.2602,"central":1.0409,"high":3.9032,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Held at the current share (prior period 2025-06-30)","periodEnd":"2025-06-30","source":{"url":"https://www1.hkexnews.hk/listedco/listconews/sehk/2026/0826/2026082600680.pdf","title":"MiniMax Group Inc. interim results announcement for the six months ended June 30, 2026","publisher":"MiniMax Group (HKEXnews)","publishedDate":"2026-08-26","accessedDate":"2026-10-22","tier":"T5","locator":"Management discussion: H1 2026 revenue US$116.6m; Open Platform and enterprise services US$73.9m; AI-native products (Hailuo AI and others) US$42.6m; no split by modality"},"note":"5LayerCake assumption (T5): no figure for the earlier period exists for this undisclosed allowance, so the share is held at the current share (the company is assumed to have grown with this part of the market)."},"image-video":{"low":0.3218,"central":0.9653,"high":2.2524,"unit":"percent","basis":"Half-year 2025 revenue × 2 in US dollars, not the card's latest-quarter × 4 run-rate","offBasis":true,"period":"H1 2025 AI-native product revenue × 2, share counted","periodEnd":"2025-06-30","source":{"url":"https://www1.hkexnews.hk/listedco/listconews/sehk/2026/0826/2026082600680.pdf","title":"MiniMax Group Inc. interim results announcement for the six months ended June 30, 2026","publisher":"MiniMax Group (HKEXnews)","publishedDate":"2026-08-26","accessedDate":"2026-10-22","tier":"T5","locator":"Financial Review, revenue by nature: AI-native products US$21,223K in the six months ended 30 June 2025 (US$42,644K in H1 2026)"},"note":"5LayerCake estimate (T5) from a filing (T1), built the same way as the current figure: AI-native products revenue of US$21.2M in H1 2025 × 2 = US$42.4M a year, of which Hailuo video is assumed to be 12-80% (central 35%). The mix of the AI-native products (Hailuo and other products such as Talkie) may have shifted since, which this does not capture; each prior value is the current value × 21.223 ÷ 42.644 = 0.4977. Prior share = the prior US$ figure ÷ the prior part total of US$1,547M, which is the current part total of US$3,758M ÷ 2.43, the growth of this part over the year: pooled from every company in this sublayer's file set with a sourced pair, because the image-and-video pairs alone (Kling US$500M ÷ US$138M, Meshy at 50% US$50M ÷ US$7.5M, MiniMax's video share US$30M ÷ US$15M) cover only 15% of this part: the four speech pairs (ElevenLabs US$600M ÷ US$200M, AssemblyAI US$76M ÷ US$26M, SoundHound US$117.7M ÷ US$94.2M, iFLYTEK US$83.5M ÷ US$70.8M) and the two avatar pairs (HeyGen US$205M ÷ US$95M, Synthesia US$150M ÷ US$100M) plus these, US$1,812M ÷ US$747M = 2.43. Companies without a sourced prior are assumed to have grown at that rate. Low and high hold the denominator at its central."},"avatar-video":{"low":0,"central":0.1005,"high":1.0054,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Held at the current share (prior period 2025-06-30)","periodEnd":"2025-06-30","source":{"url":"https://www1.hkexnews.hk/listedco/listconews/sehk/2026/0826/2026082600680.pdf","title":"MiniMax Group Inc. interim results announcement for the six months ended June 30, 2026","publisher":"MiniMax Group (HKEXnews)","publishedDate":"2026-08-26","accessedDate":"2026-10-22","tier":"T5","locator":"Management discussion: H1 2026 revenue US$116.6m; Open Platform and enterprise services US$73.9m; AI-native products (Hailuo AI and others) US$42.6m; no split by modality"},"note":"5LayerCake assumption (T5): no figure for the earlier period exists for this undisclosed allowance, so the share is held at the current share (the company is assumed to have grown with this part of the market)."},"robotics-science":{"low":0,"central":0.1284,"high":1.2837,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Held at the current share (prior period 2025-06-30)","periodEnd":"2025-06-30","source":{"url":"https://www1.hkexnews.hk/listedco/listconews/sehk/2026/0826/2026082600680.pdf","title":"MiniMax Group Inc. interim results announcement for the six months ended June 30, 2026","publisher":"MiniMax Group (HKEXnews)","publishedDate":"2026-08-26","accessedDate":"2026-10-22","tier":"T5","locator":"Management discussion: H1 2026 revenue US$116.6m; Open Platform and enterprise services US$73.9m; AI-native products (Hailuo AI and others) US$42.6m; no split by modality"},"note":"5LayerCake assumption (T5): no figure for the earlier period exists for this undisclosed allowance, so the share is held at the current share (the company is assumed to have grown with this part of the market)."}},"activityGrowth":{"current":{"low":42644000,"central":42644000,"high":42644000,"unit":"USD","basis":"AI-native products revenue (Hailuo AI and other products), half-year, US dollars (unaudited; proxy for specialist-model product revenue)","period":"H1 2026 (half-year)","periodEnd":"2026-06-30","source":{"url":"https://www1.hkexnews.hk/listedco/listconews/sehk/2026/0826/2026082600680.pdf","title":"MiniMax Group Inc. interim results announcement for the six months ended June 30, 2026","publisher":"MiniMax Group (HKEXnews)","publishedDate":"2026-08-26","accessedDate":"2026-10-22","tier":"T1","locator":"Financial Review, revenue by nature: AI-native products US$42,644K (36.6% of revenue) for the six months ended 30 June 2026"},"note":"Proxy: the line holds Hailuo AI (video) and other AI-native products, including some that count in Applications. Open Platform revenue (US$73.9M, +703%) is mostly language-model tokens and is not used."},"prior":{"low":21223000,"central":21223000,"high":21223000,"unit":"USD","basis":"AI-native products revenue (Hailuo AI and other products), half-year, US dollars (unaudited; proxy for specialist-model product revenue)","period":"H1 2025 (half-year)","periodEnd":"2025-06-30","source":{"url":"https://www1.hkexnews.hk/listedco/listconews/sehk/2026/0826/2026082600680.pdf","title":"MiniMax Group Inc. interim results announcement for the six months ended June 30, 2026","publisher":"MiniMax Group (HKEXnews)","publishedDate":"2026-08-26","accessedDate":"2026-10-22","tier":"T1","locator":"Financial Review, revenue by nature, prior-period column: AI-native products US$21,223K (69.7% of revenue)"},"note":"Same table, comparative column. Growth 42.644 ÷ 21.223 = 2.0 times (the filing says +100.9%)."}},"screening":{"speech-voice":{"levelLow":1,"levelHigh":2},"music":{"levelLow":1,"levelHigh":2},"image-video":{"levelLow":1,"levelHigh":2},"avatar-video":{"levelLow":1,"levelHigh":2},"robotics-science":{"levelLow":1,"levelHigh":2}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"6458ae59539282988cbb90d50f188927f0d031c6d2bd615b67fe2ffb0b803273"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"6458ae59539282988cbb90d50f188927f0d031c6d2bd615b67fe2ffb0b803273"}}},"stability-ai":{"schema":"5lc-index-inputs/1","key":"models-specialist-models","entityId":"stability-ai","name":"Stability AI","kind":"private","commercialActivityVerified":true,"shares":{"speech-voice":{"participates":false,"evidence":{"url":"https://sacra.com/c/stability-ai/","title":"Stability AI (Sacra company profile)","publisher":"Sacra","publishedDate":"undated","accessedDate":"2026-10-22","tier":"T4","locator":"Sacra describes Stability AI's image, video, 3D and audio models; no speech, avatar or robot models."}},"music":{"low":0,"central":0.5204,"high":2.6021,"unit":"percent","basis":"Calendar-year 2024 revenue in US dollars (Sacra estimate), not the card's annualised run-rate","offBasis":true,"period":"CY2024 estimate (Stable Audio)","periodEnd":"2024-12-31","source":{"url":"https://sacra.com/c/stability-ai/","title":"Stability AI (Sacra company profile)","publisher":"Sacra","publishedDate":"undated","accessedDate":"2026-10-22","tier":"T5","locator":"Sacra estimates about $50M revenue in 2024 (up from about $8M in 2023); mainly API credits and enterprise licensing; Stable Audio mentioned without figures"},"note":"5LayerCake estimate (T5). Sacra mentions Stable Audio and music-label partnerships without figures. Allowance of up to 20% of its revenue. Share = the central US$ figure ÷ the part total of US$384M, which is the register candidates' central figures (US$364M) plus an outside-register allowance of US$20M for other music-model developers not in the register (for example Producer.ai, AIVA, Soundraw); the card's own estimate for this part is US$315M-336M-500M (low-central-high). Low and high shares hold the denominator at its central."},"image-video":{"low":0.6652,"central":1.1175,"high":2.3947,"unit":"percent","basis":"Calendar-year 2024 revenue in US dollars (Sacra estimate), not the card's annualised run-rate","offBasis":true,"period":"CY2024 revenue, image portion (estimate)","periodEnd":"2024-12-31","source":{"url":"https://sacra.com/c/stability-ai/","title":"Stability AI (Sacra company profile)","publisher":"Sacra","publishedDate":"undated","accessedDate":"2026-10-22","tier":"T5","locator":"Sacra estimates about $50M revenue in 2024 (up from about $8M in 2023); mainly API credits and enterprise licensing; Stable Audio mentioned without figures"},"note":"5LayerCake estimate (T5). Sacra estimates 2024 revenue of about $50M (mainly API credits and enterprise licensing for image models, plus video and 3D). Image, video and 3D assumed 85% (central $42M); the high end allows growth since 2024 (a $76M Series B in August 2026 followed). The date is more than 12 months old, so the engine widens it. Share = the central US$ figure ÷ the part total of US$3,758M, which is the register candidates' central figures (US$3,608M) plus an outside-register allowance of US$150M for other image and video developers not in the register (for example Higgsfield, Freepik, Tripo, Genmo); the card's own estimate for this part is US$2,140M-3,590M-5,660M (low-central-high). Low and high shares hold the denominator at its central."},"avatar-video":{"participates":false,"evidence":{"url":"https://sacra.com/c/stability-ai/","title":"Stability AI (Sacra company profile)","publisher":"Sacra","publishedDate":"undated","accessedDate":"2026-10-22","tier":"T4","locator":"Sacra describes Stability AI's image, video, 3D and audio models; no speech, avatar or robot models."}},"robotics-science":{"participates":false,"evidence":{"url":"https://sacra.com/c/stability-ai/","title":"Stability AI (Sacra company profile)","publisher":"Sacra","publishedDate":"undated","accessedDate":"2026-10-22","tier":"T4","locator":"Sacra describes Stability AI's image, video, 3D and audio models; no speech, avatar or robot models."}}},"baselineRevenueUsd":{"low":50000000,"central":50000000,"high":50000000,"unit":"USD","basis":"Revenue, calendar year 2024, US dollars (Sacra estimate)","period":"CY2024 revenue (Sacra estimate)","periodEnd":"2024-12-31","source":{"url":"https://sacra.com/c/stability-ai/","title":"Stability AI (Sacra company profile)","publisher":"Sacra","publishedDate":"undated","accessedDate":"2026-10-22","tier":"T4","locator":"Sacra estimates Stability AI's 2024 revenue at about US$50M"},"note":"Sacra estimate (T4); more than 12 months old, so the engine widens it. No 2025 or 2026 figure was found."},"screening":{"music":{"levelLow":1,"levelHigh":2},"image-video":{"levelLow":1,"levelHigh":2}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"df3e10dc502ff02347d02ad07abe275244a8ac0d1c4504ae9566de0efeca7683"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"df3e10dc502ff02347d02ad07abe275244a8ac0d1c4504ae9566de0efeca7683"}}},"mirage-captions":{"schema":"5lc-index-inputs/1","key":"models-specialist-models","entityId":"mirage-captions","name":"Mirage (Captions)","kind":"private","commercialActivityVerified":true,"shares":{"speech-voice":{"participates":false,"evidence":{"url":"https://trendingtopics.eu/mirage-raises-75m-to-push-ai-video-app-captions-into-asian-markets/","title":"Mirage raises $75M to push AI video app Captions into Asian markets","publisher":"Trending Topics","publishedDate":"2026-03-25","accessedDate":"2026-10-22","tier":"T3","locator":"Trending Topics describes Captions as an AI video editing and creation app with its own video models; no speech-model, music or robot products."}},"music":{"participates":false,"evidence":{"url":"https://trendingtopics.eu/mirage-raises-75m-to-push-ai-video-app-captions-into-asian-markets/","title":"Mirage raises $75M to push AI video app Captions into Asian markets","publisher":"Trending Topics","publishedDate":"2026-03-25","accessedDate":"2026-10-22","tier":"T3","locator":"Trending Topics describes Captions as an AI video editing and creation app with its own video models; no speech-model, music or robot products."}},"image-video":{"participates":false,"evidence":{"url":"https://trendingtopics.eu/mirage-raises-75m-to-push-ai-video-app-captions-into-asian-markets/","title":"Mirage raises $75M to push AI video app Captions into Asian markets","publisher":"Trending Topics","publishedDate":"2026-03-25","accessedDate":"2026-10-22","tier":"T3","locator":"Trending Topics describes Captions as an AI video editing and creation app with its own video models; no speech-model, music or robot products."}},"avatar-video":{"low":5.6304,"central":9.0489,"high":18.0977,"unit":"percent","basis":"Trailing 365-day in-app revenue in US dollars (Appfigures estimate), scaled up for other channels; not the card's annualised run-rate","offBasis":true,"period":"Trailing 365-day in-app revenue to March 2026, scaled","periodEnd":"2026-03-25","source":{"url":"https://trendingtopics.eu/mirage-raises-75m-to-push-ai-video-app-captions-into-asian-markets/","title":"Mirage raises $75M to push AI video app Captions into Asian markets","publisher":"Trending Topics","publishedDate":"2026-03-25","accessedDate":"2026-10-22","tier":"T5","locator":"Captions generated $28.4 million of in-app revenue over the past 365 days (Appfigures), 25% from the US; 20 million users; own trained video models"},"note":"5LayerCake estimate (T5) from a market-data figure (T4). Trending Topics (25 March 2026): Captions made US$28.4M of in-app revenue in the past 365 days (Appfigures), 25% from the US. Web, enterprise and API revenue are not in that figure, so central is 1.6x (US$45M) and high 3.2x. Captions is an editing app with its own models; only own-model revenue counts. Mirage (Captions) is counted here. Share = the central US$ figure ÷ the part total of US$497M, which is the register candidates' central figures (US$467M) plus an outside-register allowance of US$30M for other avatar-video developers not in the register (for example Akool, Argil, Creatify); the card's own estimate for this part is US$400M-450M-550M (low-central-high). Low and high shares hold the denominator at its central."},"robotics-science":{"participates":false,"evidence":{"url":"https://trendingtopics.eu/mirage-raises-75m-to-push-ai-video-app-captions-into-asian-markets/","title":"Mirage raises $75M to push AI video app Captions into Asian markets","publisher":"Trending Topics","publishedDate":"2026-03-25","accessedDate":"2026-10-22","tier":"T3","locator":"Trending Topics describes Captions as an AI video editing and creation app with its own video models; no speech-model, music or robot products."}}},"baselineRevenueUsd":{"low":28400000,"central":28400000,"high":28400000,"unit":"USD","basis":"In-app revenue of the Captions app, trailing 365 days, US dollars (Appfigures estimate)","period":"Trailing 365 days to March 2026","periodEnd":"2026-03-25","source":{"url":"https://trendingtopics.eu/mirage-raises-75m-to-push-ai-video-app-captions-into-asian-markets/","title":"Mirage raises $75M to push AI video app Captions into Asian markets","publisher":"Trending Topics","publishedDate":"2026-03-25","accessedDate":"2026-10-22","tier":"T4","locator":"Captions made US$28.4M of in-app revenue in the past 365 days (Appfigures), 25% from the US"},"note":"Market-data estimate (T4) of app-store revenue only; web, enterprise and API revenue are not in it."},"screening":{"avatar-video":{"levelLow":1,"levelHigh":2}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"c5e3bd4d5bee2f0ff7a24044c5adc2b18ec6a5b07afc859496704aa9b7966548"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"c5e3bd4d5bee2f0ff7a24044c5adc2b18ec6a5b07afc859496704aa9b7966548"}}},"hume-ai":{"schema":"5lc-index-inputs/1","key":"models-specialist-models","entityId":"hume-ai","name":"Hume AI","kind":"private","commercialActivityVerified":true,"shares":{"speech-voice":{"low":0.5323,"central":2.3953,"high":5.8551,"unit":"percent","basis":"Calendar-year 2026 revenue projection in US dollars (company statement), not the card's annualised run-rate","offBasis":true,"period":"Revenue run-rate, estimate for 2026","periodEnd":"2026-12-31","source":{"url":"https://techcrunch.com/2026/01/22/google-reportedly-snags-up-team-behind-ai-voice-startup-hume-ai","title":"Google reportedly snags up team behind AI voice startup Hume AI","publisher":"TechCrunch","publishedDate":"2026-01-22","accessedDate":"2026-10-22","tier":"T5","locator":"New CEO Andrew Ettinger: Hume \"is on track to bring in $100 million in revenue this year\"; Google DeepMind licence and hires; Hume keeps supplying its technology to other AI firms"},"note":"5LayerCake estimate (T5). New CEO Andrew Ettinger told Wired and TechCrunch (22 Jan 2026) that Hume is on track for $100M of revenue in 2026 (a company projection, not a run-rate), after a non-exclusive licence to Google DeepMind. Central is 45% of the projection, because licence fees may be one-off; low is 10%. Share = the central US$ figure ÷ the part total of US$1,879M, which is the register candidates' central figures (US$1,779M) plus an outside-register allowance of US$100M for other speech developers not in the register (for example Gladia, Soniox, Fish Audio, Murf, WellSaid, Rev); the card's own estimate for this part is US$1,050M-1,440M-2,280M (low-central-high). Low and high shares hold the denominator at its central."},"music":{"participates":false,"evidence":{"url":"https://techcrunch.com/2026/01/22/google-reportedly-snags-up-team-behind-ai-voice-startup-hume-ai","title":"Google reportedly snags up team behind AI voice startup Hume AI","publisher":"TechCrunch","publishedDate":"2026-01-22","accessedDate":"2026-10-22","tier":"T3","locator":"TechCrunch describes Hume as a voice model company whose remaining business supplies its technology to AI firms; no image, video, music, avatar or robot products."}},"image-video":{"participates":false,"evidence":{"url":"https://techcrunch.com/2026/01/22/google-reportedly-snags-up-team-behind-ai-voice-startup-hume-ai","title":"Google reportedly snags up team behind AI voice startup Hume AI","publisher":"TechCrunch","publishedDate":"2026-01-22","accessedDate":"2026-10-22","tier":"T3","locator":"TechCrunch describes Hume as a voice model company whose remaining business supplies its technology to AI firms; no image, video, music, avatar or robot products."}},"avatar-video":{"participates":false,"evidence":{"url":"https://techcrunch.com/2026/01/22/google-reportedly-snags-up-team-behind-ai-voice-startup-hume-ai","title":"Google reportedly snags up team behind AI voice startup Hume AI","publisher":"TechCrunch","publishedDate":"2026-01-22","accessedDate":"2026-10-22","tier":"T3","locator":"TechCrunch describes Hume as a voice model company whose remaining business supplies its technology to AI firms; no image, video, music, avatar or robot products."}},"robotics-science":{"participates":false,"evidence":{"url":"https://techcrunch.com/2026/01/22/google-reportedly-snags-up-team-behind-ai-voice-startup-hume-ai","title":"Google reportedly snags up team behind AI voice startup Hume AI","publisher":"TechCrunch","publishedDate":"2026-01-22","accessedDate":"2026-10-22","tier":"T3","locator":"TechCrunch describes Hume as a voice model company whose remaining business supplies its technology to AI firms; no image, video, music, avatar or robot products."}}},"screening":{"speech-voice":{"levelLow":1,"levelHigh":2}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"d7c0eb9a74b082d6529f0e1a9798cc124ed513a67cb6bed8bc534afdcbca5513"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"d7c0eb9a74b082d6529f0e1a9798cc124ed513a67cb6bed8bc534afdcbca5513"}}},"chai-discovery":{"schema":"5lc-index-inputs/1","key":"models-specialist-models","entityId":"chai-discovery","name":"Chai Discovery","kind":"private","commercialActivityVerified":true,"shares":{"speech-voice":{"participates":false,"evidence":{"url":"https://techcrunch.com/2026/01/16/from-openais-offices-to-a-deal-with-eli-lilly-how-chai-discovery-became-one-of-the-flashiest-names-in-ai-drug-development/","title":"From OpenAI's offices to a deal with Eli Lilly, how Chai Discovery became one of the flashiest names in AI drug development","publisher":"TechCrunch","publishedDate":"2026-01-16","accessedDate":"2026-10-22","tier":"T3","locator":"TechCrunch describes Chai's molecule-design models; no speech, music, image, video or avatar products."}},"music":{"participates":false,"evidence":{"url":"https://techcrunch.com/2026/01/16/from-openais-offices-to-a-deal-with-eli-lilly-how-chai-discovery-became-one-of-the-flashiest-names-in-ai-drug-development/","title":"From OpenAI's offices to a deal with Eli Lilly, how Chai Discovery became one of the flashiest names in AI drug development","publisher":"TechCrunch","publishedDate":"2026-01-16","accessedDate":"2026-10-22","tier":"T3","locator":"TechCrunch describes Chai's molecule-design models; no speech, music, image, video or avatar products."}},"image-video":{"participates":false,"evidence":{"url":"https://techcrunch.com/2026/01/16/from-openais-offices-to-a-deal-with-eli-lilly-how-chai-discovery-became-one-of-the-flashiest-names-in-ai-drug-development/","title":"From OpenAI's offices to a deal with Eli Lilly, how Chai Discovery became one of the flashiest names in AI drug development","publisher":"TechCrunch","publishedDate":"2026-01-16","accessedDate":"2026-10-22","tier":"T3","locator":"TechCrunch describes Chai's molecule-design models; no speech, music, image, video or avatar products."}},"avatar-video":{"participates":false,"evidence":{"url":"https://techcrunch.com/2026/01/16/from-openais-offices-to-a-deal-with-eli-lilly-how-chai-discovery-became-one-of-the-flashiest-names-in-ai-drug-development/","title":"From OpenAI's offices to a deal with Eli Lilly, how Chai Discovery became one of the flashiest names in AI drug development","publisher":"TechCrunch","publishedDate":"2026-01-16","accessedDate":"2026-10-22","tier":"T3","locator":"TechCrunch describes Chai's molecule-design models; no speech, music, image, video or avatar products."}},"robotics-science":{"low":2.1395,"central":8.558,"high":34.2319,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Software revenue, estimate for October 2026","periodEnd":"2026-10-22","source":{"url":"https://techcrunch.com/2026/01/16/from-openais-offices-to-a-deal-with-eli-lilly-how-chai-discovery-became-one-of-the-flashiest-names-in-ai-drug-development/","title":"From OpenAI's offices to a deal with Eli Lilly, how Chai Discovery became one of the flashiest names in AI drug development","publisher":"TechCrunch","publishedDate":"2026-01-16","accessedDate":"2026-10-22","tier":"T5","locator":"Eli Lilly will use Chai's software (Chai-2 antibody design); Series B Dec 2025 of $130M at $1.3B; no revenue stated"},"note":"5LayerCake estimate (T5). TechCrunch (16 Jan 2026): Eli Lilly will use Chai's software (Chai-2 antibody design); no revenue stated. Only software or model-access revenue counts, not drug-programme payments. Allowance only: early software deals with a few large pharma partners, against a US$1.3B valuation at the December 2025 round. Share = the central US$ figure ÷ the part total of US$234M, which is the register candidates' central figures (US$219M) plus an outside-register allowance of US$15M for other robot-model and science-model developers not in the register (for example Latent Labs, Periodic Labs, Absci); the card's own estimate for this part is US$140M-200M-350M (low-central-high). Low and high shares hold the denominator at its central."}},"screening":{"robotics-science":{"levelLow":1,"levelHigh":3}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"89aa36dba7dda19fe70d3bcb96d1408b37342575a644b24a9d8403bfb48f31b3"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"89aa36dba7dda19fe70d3bcb96d1408b37342575a644b24a9d8403bfb48f31b3"}}},"leonardo-ai":{"schema":"5lc-index-inputs/1","key":"models-specialist-models","entityId":"leonardo-ai","name":"Leonardo.Ai","kind":"unit","parentId":"canva","commercialActivityVerified":true,"shares":{"speech-voice":{"participates":false,"evidence":{"url":"https://techcrunch.com/2024/07/29/canva-acquires-leonardo-ai-to-boost-its-generative-ai-efforts","title":"Canva acquires Leonardo.ai to boost its generative AI efforts","publisher":"TechCrunch","publishedDate":"2024-07-29","accessedDate":"2026-10-22","tier":"T3","locator":"TechCrunch describes Leonardo.ai as image-generation tools, the Phoenix model and an API; no speech, music, avatar or robot models."}},"music":{"participates":false,"evidence":{"url":"https://techcrunch.com/2024/07/29/canva-acquires-leonardo-ai-to-boost-its-generative-ai-efforts","title":"Canva acquires Leonardo.ai to boost its generative AI efforts","publisher":"TechCrunch","publishedDate":"2024-07-29","accessedDate":"2026-10-22","tier":"T3","locator":"TechCrunch describes Leonardo.ai as image-generation tools, the Phoenix model and an API; no speech, music, avatar or robot models."}},"image-video":{"low":0.2661,"central":0.7982,"high":2.1286,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"ARR, estimate for October 2026","periodEnd":"2026-10-22","source":{"url":"https://techcrunch.com/2024/07/29/canva-acquires-leonardo-ai-to-boost-its-generative-ai-efforts","title":"Canva acquires Leonardo.ai to boost its generative AI efforts","publisher":"TechCrunch","publishedDate":"2024-07-29","accessedDate":"2026-10-22","tier":"T5","locator":"Leonardo.ai: over 19 million registered users, Phoenix model, API business; revenue not disclosed"},"note":"5LayerCake estimate (T5). Leonardo.ai (owned by Canva) had over 19M registered users and its Phoenix model and API at the July 2024 acquisition; revenue is not disclosed. Allowance sized at about 0.8% of the part. Unit figure, not Canva's total. Share = the central US$ figure ÷ the part total of US$3,758M, which is the register candidates' central figures (US$3,608M) plus an outside-register allowance of US$150M for other image and video developers not in the register (for example Higgsfield, Freepik, Tripo, Genmo); the card's own estimate for this part is US$2,140M-3,590M-5,660M (low-central-high). Low and high shares hold the denominator at its central."},"avatar-video":{"participates":false,"evidence":{"url":"https://techcrunch.com/2024/07/29/canva-acquires-leonardo-ai-to-boost-its-generative-ai-efforts","title":"Canva acquires Leonardo.ai to boost its generative AI efforts","publisher":"TechCrunch","publishedDate":"2024-07-29","accessedDate":"2026-10-22","tier":"T3","locator":"TechCrunch describes Leonardo.ai as image-generation tools, the Phoenix model and an API; no speech, music, avatar or robot models."}},"robotics-science":{"participates":false,"evidence":{"url":"https://techcrunch.com/2024/07/29/canva-acquires-leonardo-ai-to-boost-its-generative-ai-efforts","title":"Canva acquires Leonardo.ai to boost its generative AI efforts","publisher":"TechCrunch","publishedDate":"2024-07-29","accessedDate":"2026-10-22","tier":"T3","locator":"TechCrunch describes Leonardo.ai as image-generation tools, the Phoenix model and an API; no speech, music, avatar or robot models."}}},"screening":{"image-video":{"levelLow":1,"levelHigh":2}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"443ace4dd375873cb67b123f13af92b8eaf118943c600813fad81daaeabdd0ba"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"443ace4dd375873cb67b123f13af92b8eaf118943c600813fad81daaeabdd0ba"}}},"lightricks":{"schema":"5lc-index-inputs/1","key":"models-specialist-models","entityId":"lightricks","name":"Lightricks (LTX)","kind":"private","commercialActivityVerified":true,"shares":{"speech-voice":{"participates":false,"evidence":{"url":"https://www.calcalistech.com/ctechnews/article/r1dgjt5gmg","title":"Lightricks to split into two companies as it cuts another 75 jobs","publisher":"CTech by Calcalist","publishedDate":"2026","accessedDate":"2026-10-22","tier":"T3","locator":"CTech describes Lightricks' LTX video-model business and the Facetune app; no speech, music, avatar or robot models."}},"music":{"participates":false,"evidence":{"url":"https://www.calcalistech.com/ctechnews/article/r1dgjt5gmg","title":"Lightricks to split into two companies as it cuts another 75 jobs","publisher":"CTech by Calcalist","publishedDate":"2026","accessedDate":"2026-10-22","tier":"T3","locator":"CTech describes Lightricks' LTX video-model business and the Facetune app; no speech, music, avatar or robot models."}},"image-video":{"low":0.3991,"central":0.7982,"high":2.3947,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Annualised licensing revenue (LTX), estimate","periodEnd":"2026-06-01","source":{"url":"https://www.calcalistech.com/ctechnews/article/r1dgjt5gmg","title":"Lightricks to split into two companies as it cuts another 75 jobs","publisher":"CTech by Calcalist","publishedDate":"2026","accessedDate":"2026-10-22","tier":"T5","locator":"Facetune about $300 million a year; LTX is \"already generating annualized revenues of tens of millions of dollars from licensing its models\"; article dated 01.06.26 (1 June or 6 January 2026)"},"note":"5LayerCake estimate (T5) from a company statement (T2/T3). CTech reports that LTX is 'already generating annualized revenues of tens of millions of dollars from licensing its models', mainly from large tech and Fortune 500 clients; the article date is ambiguous (01.06.26: 1 June or 6 January 2026). 'Tens of millions' read as US$15-90M, central US$30M. Facetune (about US$300M a year) is an app and counts in Applications. Share = the central US$ figure ÷ the part total of US$3,758M, which is the register candidates' central figures (US$3,608M) plus an outside-register allowance of US$150M for other image and video developers not in the register (for example Higgsfield, Freepik, Tripo, Genmo); the card's own estimate for this part is US$2,140M-3,590M-5,660M (low-central-high). Low and high shares hold the denominator at its central."},"avatar-video":{"participates":false,"evidence":{"url":"https://www.calcalistech.com/ctechnews/article/r1dgjt5gmg","title":"Lightricks to split into two companies as it cuts another 75 jobs","publisher":"CTech by Calcalist","publishedDate":"2026","accessedDate":"2026-10-22","tier":"T3","locator":"CTech describes Lightricks' LTX video-model business and the Facetune app; no speech, music, avatar or robot models."}},"robotics-science":{"participates":false,"evidence":{"url":"https://www.calcalistech.com/ctechnews/article/r1dgjt5gmg","title":"Lightricks to split into two companies as it cuts another 75 jobs","publisher":"CTech by Calcalist","publishedDate":"2026","accessedDate":"2026-10-22","tier":"T3","locator":"CTech describes Lightricks' LTX video-model business and the Facetune app; no speech, music, avatar or robot models."}}},"screening":{"image-video":{"levelLow":1,"levelHigh":2}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"2a777b08cbcf121f60b0072a9c0ed00c75b962dd932d16c516d0652cf083e061"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"2a777b08cbcf121f60b0072a9c0ed00c75b962dd932d16c516d0652cf083e061"}}},"krea":{"schema":"5lc-index-inputs/1","key":"models-specialist-models","entityId":"krea","name":"Krea","kind":"private","commercialActivityVerified":true,"shares":{"speech-voice":{"participates":false,"evidence":{"url":"https://www.krea.ai/pricing","title":"Krea pricing","publisher":"Krea","publishedDate":"undated","tier":"T2","locator":"Krea's pricing page lists image, video and 3D models and lip-sync in every plan; it names no speech or voice model, avatar-video product or robot model","accessedDate":"2026-10-22"}},"music":{"low":0,"central":0.2604,"high":1.3021,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"Estimate for October 2026","periodEnd":"2026-10-22","source":{"url":"https://www.krea.ai/pricing","title":"Krea pricing","publisher":"Krea","publishedDate":"undated","tier":"T5","locator":"Plans from US$5 to US$160 a month plus Enterprise; image models (Krea 1, 2, 3 and third-party), video (Krea Realtime and third-party) and 3D; an 'Audio' navigation entry names no model and shows no price","accessedDate":"2026-10-22"},"note":"5LayerCake estimate (T5). Krea's pricing page shows an 'Audio' entry in its navigation but names no model and shows no price, and ContentGrip (April 2025) said audio and music generation were 'in the works'. Allowance for possible undisclosed revenue, not a measured figure: central US$1M, high US$5M. Share = the central US$ figure ÷ the part total of US$384M. That is the screening total (the register candidates' central figures of US$364M plus an outside-register allowance of US$20M for other developers not in the register), held fixed so every company in this sublayer is measured against one total; the card's own estimate for this part is US$315M-336M-500M (low-central-high). Low and high shares hold the denominator at its central."},"image-video":{"low":0.3725,"central":0.6652,"high":1.8627,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"ARR, estimate for October 2026","periodEnd":"2026-10-22","source":{"url":"https://contentgrip.com/krea-ai-platform-funding-genai-creatives","title":"Krea raises $83M to simplify GenAI design workflows","publisher":"ContentGrip","publishedDate":"2025-04-08","tier":"T5","locator":"US$47M Series B led by Bain Capital Ventures; total funding US$83M; 'According to sources cited by TechCrunch, Krea is now valued at US$500 million post-money'; revenue and user counts not disclosed","accessedDate":"2026-10-22"},"note":"5LayerCake estimate (T5). ContentGrip (8 April 2025): US$47M Series B at about US$500M post-money (valuation per sources cited by TechCrunch); no revenue disclosed. Krea's pricing page (opened 22 October 2026) shows paid plans of US$5-160 a month and an Enterprise tier, with Krea's own models (Krea 1, 2, 3) next to many third-party ones, so only its own models are counted. Valuation divided by a peer ratio of valuation to ARR of 12-37x gives US$14-42M; central US$25M; high US$70M allows growth since. Further inputs: Krea, 22 October 2026, https://www.krea.ai/pricing. Share = the central US$ figure ÷ the part total of US$3,758M. That is the screening total (the register candidates' central figures of US$3,608M plus an outside-register allowance of US$150M for other developers not in the register), held fixed so every company in this sublayer is measured against one total; the card's own estimate for this part is US$2,140M-3,590M-5,660M (low-central-high). Low and high shares hold the denominator at its central."},"avatar-video":{"participates":false,"evidence":{"url":"https://www.krea.ai/pricing","title":"Krea pricing","publisher":"Krea","publishedDate":"undated","tier":"T2","locator":"Krea's pricing page lists image, video and 3D models and lip-sync in every plan; it names no speech or voice model, avatar-video product or robot model","accessedDate":"2026-10-22"}},"robotics-science":{"participates":false,"evidence":{"url":"https://www.krea.ai/pricing","title":"Krea pricing","publisher":"Krea","publishedDate":"undated","tier":"T2","locator":"Krea's pricing page lists image, video and 3D models and lip-sync in every plan; it names no speech or voice model, avatar-video product or robot model","accessedDate":"2026-10-22"}}},"screening":{"music":{"levelLow":1,"levelHigh":2},"image-video":{"levelLow":1,"levelHigh":2}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"9e53f5bab9f8a247381028a546d7ea2f6ea290a2483390e7bcd3b3671beb9b93"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"9e53f5bab9f8a247381028a546d7ea2f6ea290a2483390e7bcd3b3671beb9b93"}}},"luma-ai":{"schema":"5lc-index-inputs/1","key":"models-specialist-models","entityId":"luma-ai","name":"Luma AI","kind":"private","commercialActivityVerified":true,"shares":{"speech-voice":{"participates":false,"evidence":{"url":"https://sacra.com/c/luma-ai/","title":"Luma AI (Sacra company profile)","publisher":"Sacra","publishedDate":"undated","accessedDate":"2026-10-22","tier":"T4","locator":"Sacra and SiliconANGLE describe Luma as a multimodal video-model company (Dream Machine); no speech, music, avatar or robot models."}},"music":{"participates":false,"evidence":{"url":"https://sacra.com/c/luma-ai/","title":"Luma AI (Sacra company profile)","publisher":"Sacra","publishedDate":"undated","accessedDate":"2026-10-22","tier":"T4","locator":"Sacra and SiliconANGLE describe Luma as a multimodal video-model company (Dream Machine); no speech, music, avatar or robot models."}},"image-video":{"low":0.2129,"central":0.6652,"high":3.1929,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"ARR, estimate for October 2026","periodEnd":"2026-10-22","source":{"url":"https://sacra.com/c/luma-ai/","title":"Luma AI (Sacra company profile)","publisher":"Sacra","publishedDate":"undated","accessedDate":"2026-10-22","tier":"T5","locator":"Sacra estimates about $8M annualized revenue as of December 2024; no 2025 or 2026 figure"},"note":"5LayerCake estimate (T5). Sacra estimates about US$8M annualized revenue in December 2024 and has no later figure. SiliconANGLE (19 Nov 2025): US$900M Series C led by HUMAIN at more than US$4B, with film studios, agencies and marketing teams as users. Central US$25M assumes 3x growth over two years; the high end of US$120M allows for valuation-implied scale (the peer ratio of valuation to ARR is 12-37x) and is generous. Further inputs: SiliconANGLE, 2025-11-19, https://siliconangle.com/2025/11/19/multimodal-ai-developer-luma-ai-raises-900m-funding/. Share = the central US$ figure ÷ the part total of US$3,758M, which is the register candidates' central figures (US$3,608M) plus an outside-register allowance of US$150M for other image and video developers not in the register (for example Higgsfield, Freepik, Tripo, Genmo); the card's own estimate for this part is US$2,140M-3,590M-5,660M (low-central-high). Low and high shares hold the denominator at its central."},"avatar-video":{"participates":false,"evidence":{"url":"https://sacra.com/c/luma-ai/","title":"Luma AI (Sacra company profile)","publisher":"Sacra","publishedDate":"undated","accessedDate":"2026-10-22","tier":"T4","locator":"Sacra and SiliconANGLE describe Luma as a multimodal video-model company (Dream Machine); no speech, music, avatar or robot models."}},"robotics-science":{"participates":false,"evidence":{"url":"https://sacra.com/c/luma-ai/","title":"Luma AI (Sacra company profile)","publisher":"Sacra","publishedDate":"undated","accessedDate":"2026-10-22","tier":"T4","locator":"Sacra and SiliconANGLE describe Luma as a multimodal video-model company (Dream Machine); no speech, music, avatar or robot models."}}},"baselineRevenueUsd":{"low":8000000,"central":8000000,"high":8000000,"unit":"USD","basis":"Annualised revenue of Luma AI, US dollars (Sacra estimate)","period":"Annualised revenue, December 2024 (Sacra estimate)","periodEnd":"2024-12-31","source":{"url":"https://sacra.com/c/luma-ai/","title":"Luma AI (Sacra company profile)","publisher":"Sacra","publishedDate":"undated","accessedDate":"2026-10-22","tier":"T4","locator":"Sacra estimates about US$8M of annualised revenue in December 2024; no later figure"},"note":"Sacra estimate (T4); more than 12 months old, so the engine widens it. The company has raised US$900M since (November 2025) but gives no revenue figure."},"screening":{"image-video":{"levelLow":1,"levelHigh":2}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"69618cb06ed845f06ac6151a7ef2308e53adc51c97934426b3c1bf6158077627"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"69618cb06ed845f06ac6151a7ef2308e53adc51c97934426b3c1bf6158077627"}}},"shengshu":{"schema":"5lc-index-inputs/1","key":"models-specialist-models","entityId":"shengshu","name":"Shengshu (Vidu)","kind":"private","commercialActivityVerified":true,"shares":{"speech-voice":{"participates":false,"evidence":{"url":"https://kr-asia.com/chinas-video-ai-startups-are-doing-what-llm-firms-havent-making-money","title":"China's video AI startups are doing what LLM firms haven't: Making money","publisher":"KrASIA (from 36Kr)","publishedDate":"2025-07-11","accessedDate":"2026-10-22","tier":"T3","locator":"KrASIA describes Shengshu's Vidu as a video generation product; no speech, music, avatar or robot models."}},"music":{"participates":false,"evidence":{"url":"https://kr-asia.com/chinas-video-ai-startups-are-doing-what-llm-firms-havent-making-money","title":"China's video AI startups are doing what LLM firms haven't: Making money","publisher":"KrASIA (from 36Kr)","publishedDate":"2025-07-11","accessedDate":"2026-10-22","tier":"T3","locator":"KrASIA describes Shengshu's Vidu as a video generation product; no speech, music, avatar or robot models."}},"image-video":{"low":0.2661,"central":0.6652,"high":1.3305,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"ARR, estimate for October 2026","periodEnd":"2026-10-22","source":{"url":"https://kr-asia.com/chinas-video-ai-startups-are-doing-what-llm-firms-havent-making-money","title":"China's video AI startups are doing what LLM firms haven't: Making money","publisher":"KrASIA (from 36Kr)","publishedDate":"2025-07-11","accessedDate":"2026-10-22","tier":"T5","locator":"Vidu and Hailuo AI each above US$10m ARR (web only); PixVerse subscription revenue covers most costs; Kling US$100m ARR (June 2025)"},"note":"5LayerCake estimate (T5). KrASIA/36Kr (11 July 2025): Vidu above US$10M ARR on its web product, with insiders saying subscription revenue may be higher. Growth since then assumed 1x-5x (central 2.5x) in the video-model market, a judgment: US$10M × 1 / 2.5 / 5 = US$10M / 25M / 50M, which are 0.2661% / 0.6652% / 1.3305% of the part total below. Two later reports give no usable base but sit inside this range: 36Kr (Zhidx, 5 February 2026, https://eu.36kr.com/en/p/3670111938225032, opened) says Shengshu's users and revenue grew more than tenfold in 2025, with no amount; 36Kr (ZAKER Finance, 20 August 2026, https://eu.36kr.com/de/p/3947480095430272, opened) says Vidu's ARR exceeded US$20M (about RMB140M) 8 months after its market launch, without naming the launch or product (T3, press, not used as a value). Share = the central US$ figure ÷ the part total of US$3,758M, which is the register candidates' central figures (US$3,608M) plus an outside-register allowance of US$150M for other image and video developers not in the register (for example Higgsfield, Freepik, Tripo, Genmo); the card's own estimate for this part is US$2,140M-3,590M-5,660M (low-central-high). Low and high shares hold the denominator at its central."},"avatar-video":{"participates":false,"evidence":{"url":"https://kr-asia.com/chinas-video-ai-startups-are-doing-what-llm-firms-havent-making-money","title":"China's video AI startups are doing what LLM firms haven't: Making money","publisher":"KrASIA (from 36Kr)","publishedDate":"2025-07-11","accessedDate":"2026-10-22","tier":"T3","locator":"KrASIA describes Shengshu's Vidu as a video generation product; no speech, music, avatar or robot models."}},"robotics-science":{"participates":false,"evidence":{"url":"https://kr-asia.com/chinas-video-ai-startups-are-doing-what-llm-firms-havent-making-money","title":"China's video AI startups are doing what LLM firms haven't: Making money","publisher":"KrASIA (from 36Kr)","publishedDate":"2025-07-11","accessedDate":"2026-10-22","tier":"T3","locator":"KrASIA describes Shengshu's Vidu as a video generation product; no speech, music, avatar or robot models."}}},"baselineRevenueUsd":{"low":10000000,"central":10000000,"high":10000000,"unit":"USD","basis":"Annual recurring revenue (ARR) of Vidu (web product), US dollars (press report)","period":"ARR, mid-2025 (reported July 2025)","periodEnd":"2025-07-11","source":{"url":"https://kr-asia.com/chinas-video-ai-startups-are-doing-what-llm-firms-havent-making-money","title":"China's video AI startups are doing what LLM firms haven't: Making money","publisher":"KrASIA (from 36Kr)","publishedDate":"2025-07-11","accessedDate":"2026-10-22","tier":"T3","locator":"KrASIA/36Kr (11 July 2025): Vidu ARR above US$10M on its web product"},"note":"'Above US$10M': a floor, and more than 12 months old, so the engine widens it. 36Kr reported on 5 February 2026 (https://eu.36kr.com/en/p/3670111938225032) that Shengshu's users and revenue grew more than tenfold in 2025, and on 20 August 2026 (ZAKER Finance, https://eu.36kr.com/de/p/3947480095430272) a Vidu ARR above about US$20M eight months after launch; neither gives a dated base that could serve as a prior, so neither replaces this value."},"screening":{"image-video":{"levelLow":1,"levelHigh":2}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"a4eddc6ca429b0c78af6658f10ffc317fdb4743a40c1d4f762e215daee8bb3cf"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"a4eddc6ca429b0c78af6658f10ffc317fdb4743a40c1d4f762e215daee8bb3cf"}}},"cartesia":{"schema":"5lc-index-inputs/1","key":"models-specialist-models","entityId":"cartesia","name":"Cartesia","kind":"private","commercialActivityVerified":true,"shares":{"speech-voice":{"low":0.4258,"central":1.5968,"high":4.7905,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"ARR, estimate for October 2026","periodEnd":"2026-10-22","source":{"url":"https://sacra.com/c/cartesia/","title":"Cartesia (Sacra company profile)","publisher":"Sacra","publishedDate":"undated","accessedDate":"2026-10-22","tier":"T5","locator":"No revenue estimate; total funding $191M ($27M seed Dec 2024, $64M Series A Mar 2025, $100M Series A-II Oct 2025)"},"note":"5LayerCake estimate (T5). No revenue disclosed. Method: capital raised $191M (Sacra) times an ARR-to-capital ratio of 0.04-0.47; the top ratio is Deepgram's ($100M ÷ about $216M raised = 0.46), the central 0.16 reflects a company that launched Sonic in 2024. Cartesia's pricing page shows paid plans from $5 to $299 a month plus enterprise, and its customers page names about 40 customers (company-stated volumes such as 500,000+ daily voice calls at Bolna). Further inputs: Cartesia, undated, https://www.cartesia.ai/pricing; Cartesia, undated, https://www.cartesia.ai/customers; Sacra, undated, https://sacra.com/c/deepgram/. Share = the central US$ figure ÷ the part total of US$1,879M, which is the register candidates' central figures (US$1,779M) plus an outside-register allowance of US$100M for other speech developers not in the register (for example Gladia, Soniox, Fish Audio, Murf, WellSaid, Rev); the card's own estimate for this part is US$1,050M-1,440M-2,280M (low-central-high). Low and high shares hold the denominator at its central."},"music":{"participates":false,"evidence":{"url":"https://www.cartesia.ai/pricing","title":"Cartesia pricing","publisher":"Cartesia","publishedDate":"undated","accessedDate":"2026-10-22","tier":"T2","locator":"Cartesia's pricing page lists Sonic (text to speech), Ink (speech to text) and managed voice agents only."}},"image-video":{"participates":false,"evidence":{"url":"https://www.cartesia.ai/pricing","title":"Cartesia pricing","publisher":"Cartesia","publishedDate":"undated","accessedDate":"2026-10-22","tier":"T2","locator":"Cartesia's pricing page lists Sonic (text to speech), Ink (speech to text) and managed voice agents only."}},"avatar-video":{"participates":false,"evidence":{"url":"https://www.cartesia.ai/pricing","title":"Cartesia pricing","publisher":"Cartesia","publishedDate":"undated","accessedDate":"2026-10-22","tier":"T2","locator":"Cartesia's pricing page lists Sonic (text to speech), Ink (speech to text) and managed voice agents only."}},"robotics-science":{"participates":false,"evidence":{"url":"https://www.cartesia.ai/pricing","title":"Cartesia pricing","publisher":"Cartesia","publishedDate":"undated","accessedDate":"2026-10-22","tier":"T2","locator":"Cartesia's pricing page lists Sonic (text to speech), Ink (speech to text) and managed voice agents only."}}},"screening":{"speech-voice":{"levelLow":1,"levelHigh":2}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"eec4df0a3b67d99d1c1aacf82d2218a967d25a409273cafd8f11c282c805a50e"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"eec4df0a3b67d99d1c1aacf82d2218a967d25a409273cafd8f11c282c805a50e"}}},"decart":{"schema":"5lc-index-inputs/1","key":"models-specialist-models","entityId":"decart","name":"Decart","kind":"private","commercialActivityVerified":true,"shares":{"speech-voice":{"participates":false,"evidence":{"url":"https://pulse2.com/decart-300-million-raised-at-nearly-4-billion-valuation-raised-to-make-switching-ai-chips-easier/","title":"Decart: $300 Million Raised At Nearly $4 Billion Valuation Raised To Make Switching AI Chips Easier","publisher":"Pulse 2.0","publishedDate":"2026-05-19","accessedDate":"2026-10-22","tier":"T3","locator":"Pulse 2.0 describes Decart's Lucy real-time models, optimisation stack and Oasis simulation; no speech, music or avatar products. Oasis targets robotics and autonomous-system simulation, not robot control."}},"music":{"participates":false,"evidence":{"url":"https://pulse2.com/decart-300-million-raised-at-nearly-4-billion-valuation-raised-to-make-switching-ai-chips-easier/","title":"Decart: $300 Million Raised At Nearly $4 Billion Valuation Raised To Make Switching AI Chips Easier","publisher":"Pulse 2.0","publishedDate":"2026-05-19","accessedDate":"2026-10-22","tier":"T3","locator":"Pulse 2.0 describes Decart's Lucy real-time models, optimisation stack and Oasis simulation; no speech, music or avatar products. Oasis targets robotics and autonomous-system simulation, not robot control."}},"image-video":{"low":0.133,"central":0.5322,"high":2.1286,"unit":"percent","basis":"Share of the function's annualised revenue run-rate in US dollars from the company's own specialist models, through its apps, API and licences (latest month × 12 or latest quarter × 4)","period":"ARR, estimate for October 2026","periodEnd":"2026-10-22","source":{"url":"https://pulse2.com/decart-300-million-raised-at-nearly-4-billion-valuation-raised-to-make-switching-ai-chips-easier/","title":"Decart: $300 Million Raised At Nearly $4 Billion Valuation Raised To Make Switching AI Chips Easier","publisher":"Pulse 2.0","publishedDate":"2026-05-19","accessedDate":"2026-10-22","tier":"T5","locator":"Lucy real-time world model deployed with enterprise customers; AWS commercial agreement; customer API and self-serve platform; revenue not disclosed"},"note":"5LayerCake estimate (T5). Pulse 2.0 (19 May 2026): US$300M round at nearly US$4B; Lucy is deployed with enterprise customers and there is an AWS commercial agreement; revenue is not disclosed and part of it is optimisation software, not models. Allowance only. Share = the central US$ figure ÷ the part total of US$3,758M, which is the register candidates' central figures (US$3,608M) plus an outside-register allowance of US$150M for other image and video developers not in the register (for example Higgsfield, Freepik, Tripo, Genmo); the card's own estimate for this part is US$2,140M-3,590M-5,660M (low-central-high). Low and high shares hold the denominator at its central."},"avatar-video":{"participates":false,"evidence":{"url":"https://pulse2.com/decart-300-million-raised-at-nearly-4-billion-valuation-raised-to-make-switching-ai-chips-easier/","title":"Decart: $300 Million Raised At Nearly $4 Billion Valuation Raised To Make Switching AI Chips Easier","publisher":"Pulse 2.0","publishedDate":"2026-05-19","accessedDate":"2026-10-22","tier":"T3","locator":"Pulse 2.0 describes Decart's Lucy real-time models, optimisation stack and Oasis simulation; no speech, music or avatar products. Oasis targets robotics and autonomous-system simulation, not robot control."}},"robotics-science":{"participates":false,"evidence":{"url":"https://pulse2.com/decart-300-million-raised-at-nearly-4-billion-valuation-raised-to-make-switching-ai-chips-easier/","title":"Decart: $300 Million Raised At Nearly $4 Billion Valuation Raised To Make Switching AI Chips Easier","publisher":"Pulse 2.0","publishedDate":"2026-05-19","accessedDate":"2026-10-22","tier":"T3","locator":"Pulse 2.0 describes Decart's Lucy real-time models, optimisation stack and Oasis simulation; no speech, music or avatar products. Oasis targets robotics and autonomous-system simulation, not robot control."}}},"screening":{"image-video":{"levelLow":1,"levelHigh":2}},"checks":{"sourceSupport":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"039ee76eeccbe9c15352a91dc530495c61ef1567c1fb590715b651c95c866c34"},"consistency":{"by":"Claude check pass, Sonnet 5.5","date":"2026-10-23","passed":true,"sha256":"039ee76eeccbe9c15352a91dc530495c61ef1567c1fb590715b651c95c866c34"}}}},"ratings":{"bytedance-seed":{"a":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"bytedance-seed","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-23","functions":{"speech-voice":{"alternatives":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/speech-to-text","date":"undated","support":"Chinese rivals Alibaba (1st, 1.7% word error) and StepFun (8th) plus global developers lead speech recognition. Page shows no publication date (date form: undated); read 2026-10-23."},{"url":"https://artificialanalysis.ai/text-to-speech/leaderboard","date":"undated","support":"Alibaba Qwen TTS (3rd) and StepFun rank in the TTS arena top 15 alongside ElevenLabs, Cartesia and Google. Page shows no publication date (date form: undated); read 2026-10-23."}]},"switchingTime":{"level":1,"evidence":[{"url":"https://livekit.com/blog/introducing-livekit-inference","date":"2025-10-01","support":"LiveKit says swapping a voice in an agent pipeline is as easy as changing the TTS string; STT and TTS from several vendors behind one key."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/text-to-speech/leaderboard","date":"undated","support":"No ByteDance voice model appears in the TTS top 15 while other Chinese and global developers do: no gap. Page shows no publication date (date form: undated); read 2026-10-23."}]}},"music":{"alternatives":{"level":1,"evidence":[{"url":"https://techcrunch.com/2026/02/27/ai-music-generator-suno-hits-2-million-paid-subscribers-and-300m-in-annual-recurring-revenue/","date":"2026-02-27","support":"Suno, the largest music-model supplier, reports 2m paid subscribers and $300m ARR"},{"url":"https://decrypt.co/358654/google-ai-music-gemini-lyria-3-review-suno-udio","date":"2026-02-19","support":"Google's Lyria 3 joins Suno and Udio as consumer music generators: many suppliers"}]},"switchingTime":{"level":1,"evidence":[{"url":"https://www.wsls.com/business/2025/11/02/ai-song-generator-udio-offers-brief-window-for-downloads-after-universal-settlement-upsets-users/","date":"2025-11-02","support":"AP reports an apparent exodus of paying Udio users within days of a download ban: music users switch fast"}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://suno.com/blog/introducing-v6","date":"2026-09-09","support":"Suno says its licensed v6 models are faster and higher quality; a larger alternative exists, no gap"}]}},"image-video":{"alternatives":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/video/leaderboard/text-to-video","date":"undated","support":"Seedance 2.5 ranks 5th (Elo 1158); Alibaba Wan 3.0 (1175), Utopai, HeyGen and MiniMax H3 are within 17 points and Google, BFL follow. Page shows no publication date (date form: undated); read 2026-10-23."},{"url":"https://techcrunch.com/2026/07/13/video-generation-startup-pixverse-raises-439m-valuation-soars-past-2b/","date":"2026-07-13","support":"TechCrunch lists Seedance, Video Rebirth, Kling, Midjourney, Runway and Luma as video rivals"}]},"switchingTime":{"level":1,"evidence":[{"url":"https://vercel.com/blog/ai-gateway-production-index-september-2026","date":"2026-09-17","support":"Seedance led video spend on Vercel's gateway but Veo rose from 15% to 20% in a month and Grok Imagine fell 42% to 19%: share moves within weeks"}]},"capabilityGap":{"level":2,"evidence":[{"url":"https://eu.36kr.com/en/p/3869247060677636","date":"2026-06-26","support":"Seedance 2.0 is Volcano Engine's main revenue driver, reportedly over 1bn yuan a month; replacing this largest share needs rivals to add capacity: small gap"},{"url":"https://artificialanalysis.ai/video/leaderboard/text-to-video","date":"undated","support":"Rivals' best video models score equal to or above Seedance, so the gap is capacity, not quality. Page shows no publication date (date form: undated); read 2026-10-23."}]}}}},"b":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"bytedance-seed","rater":"b","model":"claude-sonnet-5-5","ratedAt":"2026-10-23","functions":{"speech-voice":{"alternatives":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/text-to-speech/leaderboard","date":"undated","support":"None of the top 20 rows in the live TTS arena is a ByteDance model; Alibaba, Google, Cartesia, Inworld, Fish Audio and others fill them (page undated)."},{"url":"https://yicaiglobal.com/news/bytedances-volcano-engine-starts-price-war-in-chinas-enterprise-llms-sector","date":"2024-05-16","support":"Volcano Engine launched Doubao speech recognition and synthesis models in May 2024 with sharp price cuts, typical of a crowded market."}]},"switchingTime":{"level":1,"evidence":[{"url":"https://docs.pipecat.ai/server/services/supported-services","date":"undated","support":"Pipecat lists more than 50 TTS services and no ByteDance entry, so integrators already route to other vendors (page undated)."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/text-to-speech/leaderboard","date":"undated","support":"The top rows are held by other vendors; no ByteDance model reaches them, so the best alternative is ahead (live page, no date)."}]}},"music":{"alternatives":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/music/leaderboard/vocals","date":"undated","support":"No ByteDance model appears among the 13 vocal-arena models, which include Suno, Mureka, Google, ElevenLabs and MiniMax (live page, no date). Evidence of a product is thin."}]},"switchingTime":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/music/leaderboard/instrumental","date":"undated","support":"The instrumental arena lists seven vendors (Suno, Mureka, Google, ElevenLabs, Stability, MiniMax, Sonauto), none ByteDance; users move between such tools freely (live page, no date)."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/music/leaderboard/vocals","date":"undated","support":"Suno v6 scores 1147 and Mureka V9.5 1106; the best alternatives are ahead of any ByteDance entry (live page, no date)."}]}},"image-video":{"alternatives":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/video/leaderboard/text-to-video","date":"undated","support":"Live text-to-video arena: Seedance 2.5 scores 1158 (rank 5); Alibaba Wan 3.0 1175, Utopai X 1167, HeyGen Video 1 1160, MiniMax H3 1158 and FLUX 3 1145 are within 30 Elo (page undated)."},{"url":"https://artificialanalysis.ai/image/leaderboard/text-to-image","date":"undated","support":"Seedream 5.0 Pro scores 1100 (rank 16) on the text-to-image arena; OpenAI, Google, Microsoft, xAI, Meta and Black Forest Labs rank above (page undated)."}]},"switchingTime":{"level":2,"evidence":[{"url":"https://kr-asia.com/bytedance-raises-volcano-engines-maas-revenue-target-on-seedance-2-0-growth","date":"2026-06-05","support":"Sources told 36Kr Seedance 2.0 earns over RMB1bn a month, with about 95% penetration in short dramas; some roughly ten-second videos took up to ten hours at the Lunar New Year peak (5 Jun 2026). Rivals need to ramp capacity for that volume; level 2 is a judgement."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/video/leaderboard/text-to-video","date":"undated","support":"Wan 3.0 (1175) leads Seedance 2.5 (1158) by 17 Elo, within noise, and Google's Veo is reported to be larger by market share, so no clear capability or capacity gap (live page, no date)."},{"url":"https://kr-asia.com/bytedance-raises-volcano-engines-maas-revenue-target-on-seedance-2-0-growth","date":"2026-06-05","support":"36Kr says Seedance 2.0 ranks second globally by market share behind Google's Veo (5 Jun 2026)."}]}}}}},"elevenlabs":{"a":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"elevenlabs","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-23","functions":{"speech-voice":{"alternatives":{"level":2,"evidence":[{"url":"https://artificialanalysis.ai/text-to-speech/leaderboard","date":"undated","support":"ElevenLabs v4 leads the TTS arena (Elo 1335), but Alibaba Qwen TTS (1301), Cartesia Sonic 3.6 (1289), Google Gemini TTS (1278) and Inworld (1267) are within about 70 points: several close alternatives. Page shows no publication date (date form: undated); read 2026-10-23."},{"url":"https://artificialanalysis.ai/speech-to-speech/arena","date":"undated","support":"In the voice-agent arena Google Gemini Live, xAI Grok Voice, OpenAI GPT-Live and Cartesia rank above ElevenLabs Agents, so agent customers have several qualified alternatives. Page shows no publication date (date form: undated); read 2026-10-23."}]},"switchingTime":{"level":2,"evidence":[{"url":"https://livekit.com/blog/introducing-livekit-inference","date":"2025-10-01","support":"LiveKit says swapping a voice in an agent pipeline is as easy as changing the TTS string; STT and TTS from several vendors behind one key."},{"url":"https://techcrunch.com/2026/09/24/twenty-minutes-with-the-ceo-of-elevenlabs-now-reportedly-valued-at-22-billion/","date":"2026-09-24","support":"The company says 55%+ of its $600m ARR is classic enterprise (Klarna phone support, Deutsche Telekom, Cisco) and some customers' models were built jointly with it; re-building agents and custom voices takes months, not just a string swap"}]},"capabilityGap":{"level":2,"evidence":[{"url":"https://artificialanalysis.ai/text-to-speech/leaderboard","date":"undated","support":"Lead over the next developer is about 34 Elo points (1335 vs 1301): a small quality gap. Page shows no publication date (date form: undated); read 2026-10-23."},{"url":"https://techcrunch.com/2026/01/22/google-reportedly-snags-up-team-behind-ai-voice-startup-hume-ai","date":"2026-01-22","support":"TechCrunch says ElevenLabs crossed $330m ARR while OpenAI, Meta (Play AI) and Google push their own voice models: large rivals with capacity to absorb volume"}]}},"music":{"alternatives":{"level":1,"evidence":[{"url":"https://techcrunch.com/2026/02/27/ai-music-generator-suno-hits-2-million-paid-subscribers-and-300m-in-annual-recurring-revenue/","date":"2026-02-27","support":"Suno alone reports 2 million paid subscribers and $300m ARR, far larger than ElevenLabs' music line"},{"url":"https://decrypt.co/358654/google-ai-music-gemini-lyria-3-review-suno-udio","date":"2026-02-19","support":"Google's Lyria 3 is in Gemini; the review says Suno and Udio still lead on longer songs and controls: several music-model suppliers exist"}]},"switchingTime":{"level":1,"evidence":[{"url":"https://www.wsls.com/business/2025/11/02/ai-song-generator-udio-offers-brief-window-for-downloads-after-universal-settlement-upsets-users/","date":"2025-11-02","support":"When Udio stopped downloads after its Universal settlement, AP reports an apparent exodus among paying users: creators move between music services within days"},{"url":"https://www.musicbusinessworldwide.com/elevenlabs-officially-launches-elevenmusic-as-streaming-and-ai-creation-platform-with-licensed-remix-model/","date":"2026-05-01","support":"ElevenMusic is a consumer creation and streaming app; users sign up or leave like any subscription"}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://techcrunch.com/2026/02/27/ai-music-generator-suno-hits-2-million-paid-subscribers-and-300m-in-annual-recurring-revenue/","date":"2026-02-27","support":"Suno, the largest music-model supplier (2m paid subscribers), could absorb ElevenLabs' small music volume; no gap"},{"url":"https://suno.com/blog/introducing-v6","date":"2026-09-09","support":"The company says its v6 models, built with Warner, BMG and Believe, are its best yet: a licensed alternative at least at parity"}]}}}},"b":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"elevenlabs","rater":"b","model":"claude-sonnet-5-5","ratedAt":"2026-10-23","functions":{"speech-voice":{"alternatives":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/text-to-speech/leaderboard","date":"undated","support":"Live Speech Arena (no date): Alibaba 1301, Cartesia 1289, Google 1278, Inworld 1267, Fish Audio and Speechify all within about 90 Elo of ElevenLabs' 1335."},{"url":"https://docs.livekit.io/agents/models/tts/","date":"undated","support":"LiveKit's agent framework lists about 40 TTS providers as plugins, including Cartesia, Google, Microsoft, OpenAI, Hume and Rime (page undated)."}]},"switchingTime":{"level":2,"evidence":[{"url":"https://www.callcentrehelper.com/case-study-revolut-elevenlabs-274839.htm","date":"2026-06-17","support":"Revolut says it tested every major provider before choosing ElevenLabs, which then slotted into its existing stack through APIs; no duration is given (vendor-published case study)."},{"url":"https://techcrunch.com/2026/09/24/twenty-minutes-with-the-ceo-of-elevenlabs-now-reportedly-valued-at-22-billion/","date":"2026-09-24","support":"The CEO says 55%+ of the roughly $600M ARR is classic enterprise (company statement). No switching time is published; level 2 is a judgement for enterprise voice-agent moves needing evaluation, telephony and compliance work."}]},"capabilityGap":{"level":2,"evidence":[{"url":"https://artificialanalysis.ai/text-to-speech/leaderboard","date":"undated","support":"Eleven v4 Turbo 1335 and v4 1329 lead the live arena; the best other vendor is Alibaba at 1301, about 34 Elo lower (page undated)."},{"url":"https://techcrunch.com/2026/09/24/twenty-minutes-with-the-ceo-of-elevenlabs-now-reportedly-valued-at-22-billion/","date":"2026-09-24","support":"The CEO says the model-level quality delta is still significant but expects differences to narrow within three to five years (company claim)."}]}},"music":{"alternatives":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/music/leaderboard/vocals","date":"undated","support":"Live vocals arena (no date): Eleven Music v2.5 scores 1013; Suno v6 1147, Mureka V9.5 1106 and Google Lyria 3 Pro 1051 all rank above it, MiniMax and Sonauto follow."},{"url":"https://artificialanalysis.ai/music/leaderboard/instrumental","date":"undated","support":"Live instrumental arena (no date): Eleven Music 1021, behind Suno 1148, Mureka 1084, Lyria 1080; Stable Audio 3 Large 1014 and MiniMax 1007 are within 15 Elo."}]},"switchingTime":{"level":1,"evidence":[{"url":"https://www.aimusicpreneur.com/ai-music-news/suno-92-percent-ai-music-market-share-mid-year-2026/","date":"2026-09-15","support":"Of 1,743 released AI tracks Apr-Aug 2026, ElevenLabs appears in under 1% of workflows (secondary summary of the SIQA report), which suggests rivals are not holding a locked-in base."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/music/leaderboard/vocals","date":"undated","support":"Three vendors (Suno, Mureka, Google) outscore ElevenLabs' music model on vocals by 38 to 134 Elo, so the best alternative does not fall short (live page, no date)."}]}}}}},"suno":{"a":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"suno","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-23","functions":{"music":{"alternatives":{"level":3,"evidence":[{"url":"https://decrypt.co/358654/google-ai-music-gemini-lyria-3-review-suno-udio","date":"2026-02-19","support":"Review: Google's Lyria 3 caps songs at 30 seconds; Suno and Udio lead on longer songs and deeper controls"},{"url":"https://www.wsls.com/business/2025/11/02/ai-song-generator-udio-offers-brief-window-for-downloads-after-universal-settlement-upsets-users/","date":"2025-11-02","support":"Udio, the other full-song rival, stopped user downloads after its Universal settlement and is moving to a walled platform"},{"url":"https://www.musicbusinessworldwide.com/elevenlabs-officially-launches-elevenmusic-as-streaming-and-ai-creation-platform-with-licensed-remix-model/","date":"2026-05-01","support":"ElevenLabs launched ElevenMusic with Merlin, Kobalt and Believe licences: one licensed full-song alternative. Few at parity"}]},"switchingTime":{"level":1,"evidence":[{"url":"https://www.wsls.com/business/2025/11/02/ai-song-generator-udio-offers-brief-window-for-downloads-after-universal-settlement-upsets-users/","date":"2025-11-02","support":"AP reports an apparent exodus among paying Udio users within days of a policy change: creators switch subscriptions in weeks"},{"url":"https://techcrunch.com/2026/02/27/ai-music-generator-suno-hits-2-million-paid-subscribers-and-300m-in-annual-recurring-revenue/","date":"2026-02-27","support":"Suno sells mainly consumer subscriptions (2m paid subscribers), which can be cancelled and replaced quickly"}]},"capabilityGap":{"level":2,"evidence":[{"url":"https://techcrunch.com/2026/02/27/ai-music-generator-suno-hits-2-million-paid-subscribers-and-300m-in-annual-recurring-revenue/","date":"2026-02-27","support":"Suno reports $300m ARR and 2m paid subscribers, far more than any rival; absorbing that volume requires rivals to scale"},{"url":"https://suno.com/blog/introducing-v6","date":"2026-09-09","support":"The company says v6, made with Warner, BMG and Believe, is its best model yet; rivals publish no head-to-head tests (thin): small-to-material gap, rated small"}]}}}},"b":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"suno","rater":"b","model":"claude-sonnet-5-5","ratedAt":"2026-10-23","functions":{"music":{"alternatives":{"level":3,"evidence":[{"url":"https://artificialanalysis.ai/music/leaderboard/vocals","date":"undated","support":"Vocals arena (no date): Suno v6 1147, Mureka V9.5 1106, Google Lyria 3 Pro 1051, ElevenLabs 1013, MiniMax 1000; only Mureka and Google come within about 100 Elo."},{"url":"https://www.aimusicpreneur.com/ai-music-news/suno-92-percent-ai-music-market-share-mid-year-2026/","date":"2026-09-15","support":"Of 1,743 verified AI releases on streaming (Apr-Aug 2026), 92.8% used Suno; ElevenLabs and Udio appear in under 1% each (secondary summary of the SIQA report)."}]},"switchingTime":{"level":2,"evidence":[{"url":"https://www.musicbusinessworldwide.com/warner-music-group-settles-with-suno-strikes-first-of-its-kind-deal-with-ai-song-generator/","date":"2025-11-25","support":"Warner and Suno settled in Nov 2025; current models are to be retired for licensed ones in 2026 and downloads need a paid account. Warner and Universal also licensed Udio, so rivals need time to ramp."},{"url":"https://www.aimusicpreneur.com/ai-music-news/suno-92-percent-ai-music-market-share-mid-year-2026/","date":"2026-09-15","support":"The 92.8% share of released tracks shows habit and volume rivals must absorb; no technical lock-in is documented, so level 2 is a judgement."}]},"capabilityGap":{"level":2,"evidence":[{"url":"https://artificialanalysis.ai/music/leaderboard/vocals","date":"undated","support":"Suno v6 leads by 41 Elo over Mureka V9.5 on vocals; on instrumentals the gap to Mureka is 64 and to Lyria 3 Pro 68 (AA music arenas, no date)."}]}}}}},"google-deepmind":{"a":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"google-deepmind","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-23","functions":{"speech-voice":{"alternatives":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/speech-to-text","date":"undated","support":"Google Gemini 3.5 Transcribe ranks 9th (2.6% word error) behind Alibaba, Microsoft AI, Mistral, ElevenLabs, xAI and StepFun: many alternatives. Page shows no publication date (date form: undated); read 2026-10-23."},{"url":"https://artificialanalysis.ai/text-to-speech/leaderboard","date":"undated","support":"Gemini 3.8 Flash TTS ranks 5th (Elo 1278) behind ElevenLabs, Alibaba and Cartesia. Page shows no publication date (date form: undated); read 2026-10-23."},{"url":"https://artificialanalysis.ai/speech-to-speech/arena","date":"undated","support":"Google's Gemini Live leads the voice-agent arena, but xAI, OpenAI, Cartesia and ElevenLabs are within about 100 Elo. Page shows no publication date (date form: undated); read 2026-10-23."}]},"switchingTime":{"level":1,"evidence":[{"url":"https://livekit.com/blog/introducing-livekit-inference","date":"2025-10-01","support":"LiveKit says swapping a voice in an agent pipeline is as easy as changing the TTS string; STT and TTS from several vendors behind one key."},{"url":"https://docs.livekit.io/agents/integrations/","date":"undated","support":"LiveKit docs: STT, TTS and realtime models are AgentSession arguments; OpenAI-compatible endpoints can be swapped by base URL. Page shows no publication date (date form: undated); read 2026-10-23."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/speech-to-text","date":"undated","support":"Alibaba, Microsoft AI, Mistral, ElevenLabs and xAI all beat Google's best transcription model (2.6%) on AA word error; ElevenLabs and Alibaba top TTS: no gap. Page shows no publication date (date form: undated); read 2026-10-23."}]}},"music":{"alternatives":{"level":1,"evidence":[{"url":"https://decrypt.co/358654/google-ai-music-gemini-lyria-3-review-suno-udio","date":"2026-02-19","support":"Review of Lyria 3 in Gemini: rivals like Suno and Udio still lead with longer songs and deeper controls"},{"url":"https://techcrunch.com/2026/02/27/ai-music-generator-suno-hits-2-million-paid-subscribers-and-300m-in-annual-recurring-revenue/","date":"2026-02-27","support":"Suno reports 2 million paid subscribers and $300m ARR"}]},"switchingTime":{"level":1,"evidence":[{"url":"https://www.wsls.com/business/2025/11/02/ai-song-generator-udio-offers-brief-window-for-downloads-after-universal-settlement-upsets-users/","date":"2025-11-02","support":"AP reports an apparent exodus of paying users when Udio stopped downloads: music creators switch services within days"}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://decrypt.co/358654/google-ai-music-gemini-lyria-3-review-suno-udio","date":"2026-02-19","support":"Lyria 3 caps output at 30-second songs while Suno and Udio make longer songs: alternatives are ahead, no gap"}]}},"image-video":{"alternatives":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/video/leaderboard/text-to-video","date":"undated","support":"Gemini Omni Flash ranks 10th in text-to-video; Alibaba, Utopai, HeyGen, MiniMax, ByteDance and Black Forest Labs rank above it. Page shows no publication date (date form: undated); read 2026-10-23."},{"url":"https://artificialanalysis.ai/image/leaderboard/text-to-image","date":"undated","support":"Nano Banana 2.1 ranks 4th in text-to-image behind OpenAI; Microsoft AI, xAI, Meta, BFL, ByteDance and Alibaba follow closely. Page shows no publication date (date form: undated); read 2026-10-23."}]},"switchingTime":{"level":1,"evidence":[{"url":"https://vercel.com/blog/ai-gateway-production-index-september-2026","date":"2026-09-17","support":"On Vercel's gateway, xAI's Grok Imagine fell from 42% of videos generated in June to 19% in August: API users move between video models within weeks"},{"url":"https://www.adobe.com/products/firefly/partner-models.html","date":"undated","support":"Adobe's Firefly app offers Google, OpenAI, Black Forest Labs, Runway, Kling and Luma models side by side, so users pick another model from a menu. Page shows no publication date (date form: undated); read 2026-10-23."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://vercel.com/blog/ai-gateway-production-index-september-2026","date":"2026-09-17","support":"Nano Banana took 50% of image spend but GPT Image had 44% and led images generated; Seedance led video ahead of Veo (20%): alternatives of equal scale exist"}]}}}},"b":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"google-deepmind","rater":"b","model":"claude-sonnet-5-5","ratedAt":"2026-10-23","functions":{"speech-voice":{"alternatives":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/text-to-speech/leaderboard","date":"undated","support":"Gemini 3.8 Flash TTS scores 1278 (rank 5); ElevenLabs (1335, 1329), Alibaba (1301) and Cartesia (1289) rank above it, Inworld (1267) just below (live page, no date)."},{"url":"https://artificialanalysis.ai/speech-to-text","date":"undated","support":"Gemini 3.5 Transcribe scores 2.6% AA-WER (rank 8); six vendors score as well or better (live page, no date)."}]},"switchingTime":{"level":1,"evidence":[{"url":"https://docs.livekit.io/agents/models/tts/","date":"undated","support":"Gemini and Google Cloud are two of about 40 TTS options LiveKit supports (page undated)."},{"url":"https://developers.deepgram.com/docs/migrating-from-amazon-web-services-aws-transcribe-to-deepgram","date":"undated","support":"Deepgram's guide to leaving a large cloud STT API (AWS) is a test-and-cutover job with no stated timeline (undated)."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/text-to-speech/leaderboard","date":"undated","support":"ElevenLabs' v4 models (1335, 1329) and Alibaba (1301) outscore Google's best (1278), so the best alternative is ahead (live page, no date)."}]}},"music":{"alternatives":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/music/leaderboard/instrumental","date":"undated","support":"Lyria 3 Pro 1080 and Lyria 3.5 1078 sit with Mureka (1083-1084) behind Suno 1148; ElevenLabs, Stability and MiniMax follow at 1000-1021 (live page, no date)."}]},"switchingTime":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/music/leaderboard/vocals","date":"undated","support":"Music tools are sold as subscriptions and APIs from at least six vendors listed on the vocals arena; moving is a tool change (live page, no date)."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/music/leaderboard/vocals","date":"undated","support":"Suno v6 (1147) and Mureka V9.5 (1106) beat Lyria 3 Pro (1051), so the best alternative is ahead (live page, no date)."}]}},"image-video":{"alternatives":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/image/leaderboard/text-to-image","date":"undated","support":"Nano Banana 2.1 scores 1176 (rank 4) behind three OpenAI models; Microsoft MAI-Image-2.6 and xAI Grok Imagine 2.0 are at 1166 (page undated)."},{"url":"https://artificialanalysis.ai/video/leaderboard/text-to-video","date":"undated","support":"Gemini Omni Flash 1.1 scores 1130 (rank 10); nine models from six other vendors score higher on the video arena (page undated)."}]},"switchingTime":{"level":1,"evidence":[{"url":"https://blog.adobe.com/en/publish/2026/08/20/adobe-firefly-expands-its-creative-ai-studio-generate-music-speech-and-sound-effects-in-one-place","date":"2026-08-20","support":"Adobe's Firefly offers models from Google, Kling, Luma, OpenAI and Runway side by side (20 Aug 2026), showing apps route among suppliers per request."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/image/leaderboard/text-to-image","date":"undated","support":"OpenAI's GPT Image 2.5 (1222 and 1207) outscores Nano Banana 2.1 (1176) on images; Wan 3.0 (1175) leads video (pages undated)."}]}}}}},"kling-ai":{"a":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"kling-ai","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-23","functions":{"image-video":{"alternatives":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/video/leaderboard/text-to-video","date":"undated","support":"Kling 3.0 ranks 17th-18th (Elo 1000-1020); Alibaba, Utopai, HeyGen, MiniMax, ByteDance, BFL and Google rank above it. Page shows no publication date (date form: undated); read 2026-10-23."},{"url":"https://techcrunch.com/2026/07/13/video-generation-startup-pixverse-raises-439m-valuation-soars-past-2b/","date":"2026-07-13","support":"TechCrunch lists ByteDance Seedance, Kling, Midjourney, Runway and Luma among video rivals"}]},"switchingTime":{"level":1,"evidence":[{"url":"https://vercel.com/blog/ai-gateway-production-index-september-2026","date":"2026-09-17","support":"Vercel's gateway shows image and video share moving within weeks (Grok Imagine 42% to 19% of videos generated, June to August; Veo 15% to 20% of video spend in a month)"},{"url":"https://www.adobe.com/products/firefly/partner-models.html","date":"undated","support":"Adobe's Firefly app offers Google, OpenAI, Black Forest Labs, Runway, Kling and Luma models side by side: switching is a menu choice. Page shows no publication date (date form: undated); read 2026-10-23."},{"url":"https://www1.hkexnews.hk/listedco/listconews/sehk/2026/0819/2026081900314.pdf","date":"2026-08-19","support":"Kuaishou says Kling sells to professional film, TV and ad clients and launched MCP and CLI access for agents (p.5); still mostly self-serve and API use"}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/video/leaderboard/text-to-video","date":"undated","support":"Rival video models score up to 175 Elo above Kling 3.0, and ByteDance and Google have large capacity: no gap. Page shows no publication date (date form: undated); read 2026-10-23."},{"url":"https://www1.hkexnews.hk/listedco/listconews/sehk/2026/0819/2026081900314.pdf","date":"2026-08-19","support":"The company claims global leadership in video models (p.5), which the independent arena does not support"}]}}}},"b":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"kling-ai","rater":"b","model":"claude-sonnet-5-5","ratedAt":"2026-10-23","functions":{"image-video":{"alternatives":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/video/leaderboard/text-to-video","date":"undated","support":"Kling 3.0 Omni scores 1020 (rank 17) and Kling 3.0 1000 (rank 18); Wan 3.0, Utopai X, HeyGen, MiniMax, Seedance, FLUX 3, Gemini Omni, Agnes, Grok and others score higher (live page, no date)."}]},"switchingTime":{"level":1,"evidence":[{"url":"https://www.fool.com/earnings/call-transcripts/2026/06/11/adobe-adbe-q2-2026-earnings-transcript/","date":"2026-06-11","support":"Adobe says Firefly added partner models including Kling 3.0 and Kling 3.0 Omni (11 Jun 2026), showing tools add video suppliers quickly."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/video/leaderboard/text-to-video","date":"undated","support":"Wan 3.0 (1175) is 155 Elo above Kling 3.0 Omni (1020), so the best alternative is ahead (live page, no date)."}]}}}}},"midjourney":{"a":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"midjourney","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-23","functions":{"image-video":{"alternatives":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/image/leaderboard/text-to-image","date":"undated","support":"Text-to-image top 20 includes OpenAI, Google, Microsoft AI, xAI, Meta, BFL, ByteDance and Alibaba; Midjourney is not on the board. Page shows no publication date (date form: undated); read 2026-10-23."},{"url":"https://techcrunch.com/2026/07/13/video-generation-startup-pixverse-raises-439m-valuation-soars-past-2b/","date":"2026-07-13","support":"TechCrunch names Midjourney among many video-generation competitors with Runway, Luma, Kling and Seedance"}]},"switchingTime":{"level":1,"evidence":[{"url":"https://vercel.com/blog/ai-gateway-production-index-september-2026","date":"2026-09-17","support":"Vercel's gateway shows image and video share moving within weeks (Grok Imagine 42% to 19% of videos generated, June to August; Veo 15% to 20% of video spend in a month)"},{"url":"https://www.engadget.com/ai/meta-is-licensing-midjourneys-ai-image-and-video-tech-120012178.html","date":"2025-08-23","support":"Midjourney is sold as a public subscription service; the Meta deal licenses its tech, it does not lock in users (switching evidence thin, consumer subscriptions)"}]},"capabilityGap":{"level":2,"evidence":[{"url":"https://www.engadget.com/ai/meta-is-licensing-midjourneys-ai-image-and-video-tech-120012178.html","date":"2025-08-23","support":"Meta licensed Midjourney's aesthetic technology to improve its own models, indicating a distinctive but small quality edge"},{"url":"https://artificialanalysis.ai/image/leaderboard/text-to-image","date":"undated","support":"Several rivals score high on blind image tests, so the gap is aesthetic style and scale, rated small. Page shows no publication date (date form: undated); read 2026-10-23."}]}}}},"b":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"midjourney","rater":"b","model":"claude-sonnet-5-5","ratedAt":"2026-10-23","functions":{"image-video":{"alternatives":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/image/leaderboard/text-to-image","date":"undated","support":"No Midjourney model appears in the rows retrieved from the live text-to-image arena (top 25 plus named vendors); OpenAI, Google, Microsoft, xAI, Meta, Black Forest Labs, ByteDance and Alibaba models do (page undated)."},{"url":"https://winbuzzer.com/2026/03/31/midjourney-revenue-above-200-million-hardware-push-xcxwbn/","date":"2026-03-31","support":"Rivals named include DALL-E, Stable Diffusion and Adobe Firefly; Midjourney runs on subscriptions from $10 a month (31 Mar 2026)."}]},"switchingTime":{"level":1,"evidence":[{"url":"https://winbuzzer.com/2026/03/31/midjourney-revenue-above-200-million-hardware-push-xcxwbn/","date":"2026-03-31","support":"Paid plans start at $10 a month with no stated lock-in; Meta licensed Midjourney's technology in Aug 2025, and a new tool is a subscription away (31 Mar 2026)."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://www.gradually.ai/en/ai-models/midjourney-v8-1/","date":"undated","support":"On small one-image-per-test checks, V8.1 ranked 12th to 28th of 30 against GPT Image 2, Nano Banana and FLUX.2; evidence is thin and aesthetic preference is not measured (page undated)."}]}}}}},"adobe-firefly":{"a":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"adobe-firefly","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-23","functions":{"image-video":{"alternatives":{"level":1,"evidence":[{"url":"https://www.adobe.com/products/firefly/partner-models.html","date":"undated","support":"Adobe's own Firefly app offers Google, OpenAI, Black Forest Labs, Runway, Kling and Luma models alongside Firefly models. Page shows no publication date (date form: undated); read 2026-10-23."},{"url":"https://www.fool.com/earnings/call-transcripts/2026/06/11/adobe-adbe-q2-2026-earnings-transcript/","date":"2026-06-11","support":"Adobe says Firefly now also supports third-party Kling 3.0 and Kling 3.0 Omni"}]},"switchingTime":{"level":2,"evidence":[{"url":"https://www.fool.com/earnings/call-transcripts/2026/06/11/adobe-adbe-q2-2026-earnings-transcript/","date":"2026-06-11","support":"Adobe says Firefly enterprise is approaching $300m, with brand-safe custom models for large brands; retraining custom models and legal review elsewhere takes months"}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/image/leaderboard/text-to-image","date":"undated","support":"No Firefly model is in the text-to-image top 20; OpenAI, Google, Microsoft AI and others lead: no capability gap. Page shows no publication date (date form: undated); read 2026-10-23."}]}}}},"b":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"adobe-firefly","rater":"b","model":"claude-sonnet-5-5","ratedAt":"2026-10-23","functions":{"image-video":{"alternatives":{"level":2,"evidence":[{"url":"https://www.proskauer.com/blog/openais-copyright-shield-broadens-user-ip-indemnities-for-ai-created-content","date":"2023-11-10","support":"Adobe, Microsoft, Google and Shutterstock each offered IP indemnities for generated content in 2023 (10 Nov 2023), a few alternatives for brand-safe enterprise work; OpenAI added one too."},{"url":"https://blog.adobe.com/en/publish/2026/08/20/adobe-firefly-expands-its-creative-ai-studio-generate-music-speech-and-sound-effects-in-one-place","date":"2026-08-20","support":"Firefly itself offers models from Google, ElevenLabs, Kling, Luma, OpenAI and Runway beside its own (20 Aug 2026)."}]},"switchingTime":{"level":2,"evidence":[{"url":"https://www.fool.com/earnings/call-transcripts/2026/06/11/adobe-adbe-q2-2026-earnings-transcript/","date":"2026-06-11","support":"Adobe says Firefly ARR is approaching $300M, enterprise assets grew over 4x, named wins include Merck, SAP, Coca-Cola and Tesco, and Foundry custom models tie to an NVIDIA partnership (11 Jun 2026); embedded brand workflows take months to rebuild (judgement)."}]},"capabilityGap":{"level":2,"evidence":[{"url":"https://artificialanalysis.ai/image/leaderboard/text-to-image","date":"undated","support":"No Adobe Firefly model appears in the retrieved rows of the live text-to-image arena, so its edge is workflow and licensing rather than model score (page undated)."},{"url":"https://www.fool.com/earnings/call-transcripts/2026/06/11/adobe-adbe-q2-2026-earnings-transcript/","date":"2026-06-11","support":"Judgement: Google and OpenAI match model quality but lack Creative Cloud integration and Foundry custom models, a small shortfall (Adobe call, 11 Jun 2026)."}]}}}}},"openai":{"a":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"openai","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-23","dachiReview":true,"functions":{"speech-voice":{"alternatives":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/speech-to-text","date":"undated","support":"OpenAI's best transcription model (3.3% word error) sits just outside the top 15 behind Alibaba, Microsoft AI, Mistral, ElevenLabs, xAI and Google. Page shows no publication date (date form: undated); read 2026-10-23."},{"url":"https://artificialanalysis.ai/text-to-speech/leaderboard","date":"undated","support":"No OpenAI model is in the TTS arena top 15, which ElevenLabs, Alibaba, Cartesia, Google and Inworld lead. Page shows no publication date (date form: undated); read 2026-10-23."},{"url":"https://artificialanalysis.ai/speech-to-speech/arena","date":"undated","support":"Google Gemini Live (top 3) and xAI Grok Voice 3.0 (4th) rank above OpenAI GPT-Live-1 (5th-6th). Page shows no publication date (date form: undated); read 2026-10-23."}]},"switchingTime":{"level":1,"evidence":[{"url":"https://livekit.com/blog/introducing-livekit-inference","date":"2025-10-01","support":"LiveKit says swapping a voice in an agent pipeline is as easy as changing the TTS string; STT and TTS from several vendors behind one key."},{"url":"https://docs.livekit.io/agents/integrations/","date":"undated","support":"LiveKit docs: STT, TTS and realtime models are AgentSession arguments; OpenAI-compatible endpoints can be swapped by base URL. Page shows no publication date (date form: undated); read 2026-10-23."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/speech-to-text","date":"undated","support":"Ten-plus models beat OpenAI GPT Transcribe (3.3% word error) and Google's Gemini Live models outrank GPT-Live in the voice-agent arena: no gap. Page shows no publication date (date form: undated); read 2026-10-23."}]}},"image-video":{"alternatives":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/image/leaderboard/text-to-image","date":"undated","support":"OpenAI GPT Image 2.5 leads text-to-image (Elo 1222), with Google (1176), Microsoft AI (1166), xAI (1166) and Meta (1127) close behind. Page shows no publication date (date form: undated); read 2026-10-23."},{"url":"https://thenextweb.com/news/adobe-acquires-topaz-labs-ai-enhancement","date":"2026-06-25","support":"TNW says OpenAI shut down its Sora video app after costs ballooned; video users already rely on other suppliers"}]},"switchingTime":{"level":1,"evidence":[{"url":"https://vercel.com/blog/ai-gateway-production-index-september-2026","date":"2026-09-17","support":"Vercel's gateway shows image and video spend shifting between Google, OpenAI, ByteDance and xAI month to month"},{"url":"https://www.adobe.com/products/firefly/partner-models.html","date":"undated","support":"Adobe's Firefly app lists OpenAI GPT Image next to Google Nano Banana and FLUX as interchangeable partner models. Page shows no publication date (date form: undated); read 2026-10-23."}]},"capabilityGap":{"level":2,"evidence":[{"url":"https://artificialanalysis.ai/image/leaderboard/text-to-image","date":"undated","support":"GPT Image 2.5 leads the next developer, Google, by about 46 Elo (1222 vs 1176): a small quality lead. Page shows no publication date (date form: undated); read 2026-10-23."},{"url":"https://vercel.com/blog/ai-gateway-production-index-september-2026","date":"2026-09-17","support":"On Vercel's gateway Google's Nano Banana had 50% of image spend to GPT Image's 44%: a comparable-scale alternative exists"}]}}}},"b":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"openai","rater":"b","model":"claude-sonnet-5-5","dachiReview":true,"ratedAt":"2026-10-23","functions":{"speech-voice":{"alternatives":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/text-to-speech/leaderboard","date":"undated","support":"OpenAI TTS-1 HD scores 1103 (rank 40) and GPT-Realtime-2 1075 (rank 53); more than 30 models from ElevenLabs, Alibaba, Cartesia, Google and others score higher (live page, no date)."},{"url":"https://artificialanalysis.ai/speech-to-text","date":"undated","support":"GPT Transcribe scores 3.3% AA-WER against 1.7% for the best; at least eight vendors do better (live page, no date)."}]},"switchingTime":{"level":1,"evidence":[{"url":"https://thenextweb.com/news/openai-gpt-realtime-2-voice-models","date":"2026-05-08","support":"GPT-Realtime-2, Realtime-Translate and Realtime-Whisper are API models with named customers (Zillow, Intercom, Deutsche Telekom); The Next Web says ElevenLabs and Deepgram are the vendors most affected by the pricing (8 May 2026)."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/text-to-speech/leaderboard","date":"undated","support":"The leader scores 1335 against OpenAI's best 1103 in the live arena, so alternatives lead (no date)."}]}},"image-video":{"alternatives":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/image/leaderboard/text-to-image","date":"undated","support":"GPT Image 2.5 Sunburst leads at 1222; Google Nano Banana 2.1 1176, Microsoft MAI-Image-2.6 1166, xAI Grok Imagine 2.0 1166, Meta Muse Image 1127 and Black Forest Labs FLUX 3 1117 follow (page undated)."}]},"switchingTime":{"level":1,"evidence":[{"url":"https://blog.adobe.com/en/publish/2026/08/20/adobe-firefly-expands-its-creative-ai-studio-generate-music-speech-and-sound-effects-in-one-place","date":"2026-08-20","support":"Adobe's Firefly lists OpenAI beside Google, Kling, Luma and Runway as selectable models (20 Aug 2026), so apps can swap suppliers per request."}]},"capabilityGap":{"level":2,"evidence":[{"url":"https://artificialanalysis.ai/image/leaderboard/text-to-image","date":"undated","support":"The best non-OpenAI image model, Google's Nano Banana 2.1 (1176), scores about 46 Elo below GPT Image 2.5 Sunburst (1222) (live page, no date)."}]}}}}},"heygen":{"a":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"heygen","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-23","functions":{"avatar-video":{"alternatives":{"level":2,"evidence":[{"url":"https://docs.livekit.io/agents/integrations/avatar/","date":"undated","support":"LiveKit lists 16 avatar providers including Synthesia, D-ID, Tavus, Runway, Anam and Simli. Page shows no publication date (date form: undated); read 2026-10-23."},{"url":"https://www.synthesia.io/post/synthesia-global-expansion-austin-berlin-paris-zurich-2026","date":"2026-04-21","support":"Synthesia says it is used by over 90% of the Fortune 100; customers use it for training, onboarding and localisation: a direct enterprise alternative"}]},"switchingTime":{"level":1,"evidence":[{"url":"https://www.heygen.com/blog/heygen-surpasses-200m-arr","date":"2026-06-25","support":"The company says a digital twin needs only a 15-second video, so a customer's avatar can be recreated quickly; much use is self-serve (30m+ users)"}]},"capabilityGap":{"level":2,"evidence":[{"url":"https://artificialanalysis.ai/video/leaderboard/text-to-video","date":"undated","support":"HeyGen Video 1 ranks 3rd in text-to-video (Elo 1160), within 15 points of the leader; rivals are close in quality. Page shows no publication date (date form: undated); read 2026-10-23."},{"url":"https://www.heygen.com/blog/heygen-surpasses-200m-arr","date":"2026-06-25","support":"The company says it passed $200m ARR and serves 85% of the Fortune 100; absorbing the largest share needs rivals to scale: small gap"}]}}}},"b":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"heygen","rater":"b","model":"claude-sonnet-5-5","ratedAt":"2026-10-23","functions":{"avatar-video":{"alternatives":{"level":2,"evidence":[{"url":"https://synthesia.io/post/synthesia-global-expansion-austin-berlin-paris-zurich-2026","date":"2026-04-21","support":"Synthesia says it serves 90% of the Fortune 100 and has ISO 42001, 27001 and 27701 and SOC 2 (21 Apr 2026): a second supplier at HeyGen's scale."},{"url":"https://trendingtopics.eu/mirage-raises-75m-to-push-ai-video-app-captions-into-asian-markets/","date":"2026-03-25","support":"Mirage names D-ID and HeyGen among rivals (25 Mar 2026); HeyGen's own blog points to a list of Synthesia alternatives. A few more suppliers exist at smaller scale."}]},"switchingTime":{"level":2,"evidence":[{"url":"https://www.heygen.com/blog/heygen-surpasses-200m-arr","date":"2026-06-25","support":"HeyGen says 85% of the Fortune 100 use it for training, sales and localization, with a public API and a security portal (25 Jun 2026)."},{"url":"https://synthesia.io/post/synthesia-global-expansion-austin-berlin-paris-zurich-2026","date":"2026-04-21","support":"Synthesia reports net revenue retention above 140% and $100,000+ contracts tripling in a year (21 Apr 2026), signs that enterprise buyers stay; moves need months (judgement)."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://synthesia.io/post/synthesia-global-expansion-austin-berlin-paris-zurich-2026","date":"2026-04-21","support":"Synthesia matches HeyGen on scale and product (240+ avatars, 140+ languages, interactive avatars) (21 Apr 2026)."}]}}}}},"skild-ai":{"a":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"skild-ai","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-23","functions":{"robotics-science":{"alternatives":{"level":3,"evidence":[{"url":"https://thenextweb.com/news/fieldai-10b-europe-robot-brains","date":"2026-10-02","support":"TNW names FieldAI, Physical Intelligence (valued near $11bn), Genesis AI and Agile Robots as robot-brain rivals; only FieldAI reports comparable revenue"},{"url":"https://www.skild.ai/blogs/skild-crosses-100m-arr","date":"2026-09-10","support":"The company says ~90% of its revenue comes from manipulation across 60+ paying customers; few rivals report paid manipulation deployments (thin evidence)"}]},"switchingTime":{"level":2,"evidence":[{"url":"https://www.skild.ai/blogs/skild-crosses-100m-arr","date":"2026-09-10","support":"The company describes deployment as installation, integration, data collection and post-training per change; moving a line to another robot brain would take months"}]},"capabilityGap":{"level":3,"evidence":[{"url":"https://www.skild.ai/blogs/skild-crosses-100m-arr","date":"2026-09-10","support":"The company says it passed $100m ARR ten months after first deployment, including NVIDIA/Foxconn Blackwell assembly lines"},{"url":"https://thenextweb.com/news/fieldai-10b-europe-robot-brains","date":"2026-10-02","support":"TNW says Skild reported $100m run-rate on a stricter measure; Physical Intelligence, valued near $11bn, reports no comparable revenue: a material scale gap"}]}}}},"b":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"skild-ai","rater":"b","model":"claude-sonnet-5-5","ratedAt":"2026-10-23","functions":{"robotics-science":{"alternatives":{"level":3,"evidence":[{"url":"https://www.robotics247.com/article/nvidia-gtc-2026-nvidia-global-robotics-leaders-look-to-take-physical-ai-to-the-real-world","date":"2026-03-19","support":"NVIDIA's open GR00T N1.7 is in early access with commercial licensing, used by AGIBOT, LG, NEURA and others; N2 is only a preview (19 Mar 2026)."},{"url":"https://deepmind.google/models/gemini-robotics/","date":"undated","support":"Google's Gemini Robotics models are with 100+ trusted testers; research partners include Apptronik, Boston Dynamics and Agile Robots (page undated)."},{"url":"https://theaiinsider.tech/2026/10/14/physical-intelligence-opens-pi-api-beta-for-robot-model-training-and-deployment/","date":"2026-10-14","support":"Physical Intelligence opened a gated API beta in Oct 2026 with about twelve named early users (14 Oct 2026). Few suppliers qualify at deployment scale."}]},"switchingTime":{"level":2,"evidence":[{"url":"https://www.skild.ai/blogs/skild-crosses-100m-arr","date":"2026-09-10","support":"Skild says the first egg took a week and reliable cooking two months; it works with NVIDIA and Foxconn on Blackwell assembly and with Sumitomo Wiring (10 Sep 2026)."},{"url":"https://theaiinsider.tech/2026/10/14/physical-intelligence-opens-pi-api-beta-for-robot-model-training-and-deployment/","date":"2026-10-14","support":"Almond trained a deburring policy in four days with Physical Intelligence's API (14 Oct 2026): per-task onboarding of days to months."}]},"capabilityGap":{"level":3,"evidence":[{"url":"https://www.skild.ai/blogs/skild-crosses-100m-arr","date":"2026-09-10","support":"Skild says it has 60+ paying customers and over $100M ARR (company claim, 10 Sep 2026)."},{"url":"https://theaiinsider.tech/2026/10/14/physical-intelligence-opens-pi-api-beta-for-robot-model-training-and-deployment/","date":"2026-10-14","support":"Physical Intelligence's API is a gated beta with no stated price or general availability (14 Oct 2026); Gemini is in testers-only; no independent benchmark compares them."}]}}}}},"synthesia":{"a":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"synthesia","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-23","functions":{"avatar-video":{"alternatives":{"level":2,"evidence":[{"url":"https://www.heygen.com/blog/heygen-surpasses-200m-arr","date":"2026-06-25","support":"HeyGen says 85% of the Fortune 100 use it for training, sales enablement and localisation, Synthesia's core market"},{"url":"https://docs.livekit.io/agents/integrations/avatar/","date":"undated","support":"LiveKit lists 16 avatar providers including Synthesia, D-ID, Tavus and Runway. Page shows no publication date (date form: undated); read 2026-10-23."}]},"switchingTime":{"level":2,"evidence":[{"url":"https://www.synthesia.io/post/synthesia-global-expansion-austin-berlin-paris-zurich-2026","date":"2026-04-21","support":"The company says net revenue retention is over 140% and contracts above $100,000 tripled, with customers training distributed workforces on it: embedded enterprise workflows take months to move"}]},"capabilityGap":{"level":2,"evidence":[{"url":"https://seedcamp.com/views/synthesia-raises-200m-series-e-at-4b-valuation-to-continue-the-next-era-of-video-creation/","date":"2026-01-26","support":"Synthesia reported $150m ARR and 90% of the Fortune 100 in January 2026"},{"url":"https://www.heygen.com/blog/heygen-surpasses-200m-arr","date":"2026-06-25","support":"HeyGen reports $200m ARR, a larger rival able to absorb volume; differences are mainly enterprise controls: small gap"}]}}}},"b":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"synthesia","rater":"b","model":"claude-sonnet-5-5","ratedAt":"2026-10-23","functions":{"avatar-video":{"alternatives":{"level":2,"evidence":[{"url":"https://www.heygen.com/blog/heygen-surpasses-200m-arr","date":"2026-06-25","support":"HeyGen says 85% of the Fortune 100 use it, with Avatar V and a public API (25 Jun 2026): a second supplier at Synthesia's scale."},{"url":"https://trendingtopics.eu/mirage-raises-75m-to-push-ai-video-app-captions-into-asian-markets/","date":"2026-03-25","support":"Mirage names D-ID and HeyGen among avatar rivals (25 Mar 2026); other suppliers are smaller."}]},"switchingTime":{"level":2,"evidence":[{"url":"https://synthesia.io/post/synthesia-global-expansion-austin-berlin-paris-zurich-2026","date":"2026-04-21","support":"Synthesia says net revenue retention is over 140% and $100,000+ contracts tripled in a year; it lists ISO 42001, 27001, 27701 and SOC 2 (21 Apr 2026). Enterprise moves need months (judgement)."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://www.heygen.com/blog/heygen-surpasses-200m-arr","date":"2026-06-25","support":"HeyGen offers 85%-of-Fortune-100 use, Avatar V, translation, dubbing, API and agents (25 Jun 2026), so the best alternative matches Synthesia."},{"url":"https://artificialanalysis.ai/video/leaderboard/text-to-video","date":"undated","support":"HeyGen Video 1 ranks 3rd on the live text-to-video arena at 1160 (page undated)."}]}}}}},"runway":{"a":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"runway","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-23","functions":{"image-video":{"alternatives":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/video/leaderboard/text-to-video","date":"undated","support":"No Runway model is in the text-to-video top 20, led by Alibaba, Utopai, HeyGen, MiniMax, ByteDance, BFL and Google. Page shows no publication date (date form: undated); read 2026-10-23."},{"url":"https://www.adobe.com/products/firefly/partner-models.html","date":"undated","support":"Adobe offers Runway Gen-4.5 next to Kling, Luma and Google models in Firefly. Page shows no publication date (date form: undated); read 2026-10-23."}]},"switchingTime":{"level":1,"evidence":[{"url":"https://www.pymnts.com/?p=4167359","date":"2026-09-08","support":"The company says its largest customers increasingly license its frontier models directly; it also serves 60m+ creators on subscriptions"},{"url":"https://vercel.com/blog/ai-gateway-production-index-september-2026","date":"2026-09-17","support":"Vercel's gateway shows image and video share moving within weeks (Grok Imagine 42% to 19% of videos generated, June to August; Veo 15% to 20% of video spend in a month)"}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/video/leaderboard/text-to-video","date":"undated","support":"At least 20 video models score above Runway's on the arena: no capability gap. Page shows no publication date (date form: undated); read 2026-10-23."}]}}}},"b":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"runway","rater":"b","model":"claude-sonnet-5-5","ratedAt":"2026-10-23","functions":{"image-video":{"alternatives":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/video/leaderboard/text-to-video","date":"undated","support":"No Runway model appears among the 29 text-to-video rows retrieved (Wan, Utopai, HeyGen, MiniMax, Seedance, FLUX, Google, xAI, Kling and others do) (live page, no date)."},{"url":"https://blog.adobe.com/en/publish/2026/08/20/adobe-firefly-expands-its-creative-ai-studio-generate-music-speech-and-sound-effects-in-one-place","date":"2026-08-20","support":"Adobe lists Runway beside Google, Kling, Luma and OpenAI in Firefly (20 Aug 2026), which shows several comparable suppliers."}]},"switchingTime":{"level":1,"evidence":[{"url":"https://www.pymnts.com/?p=4167359","date":"2026-09-08","support":"Runway reports $200M ARR and says its largest customers increasingly license its models directly (8 Sep 2026); most volume is still self-serve creator subscriptions, so level 1 with upside risk."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/video/leaderboard/text-to-video","date":"undated","support":"Thirteen models from eight or more vendors score above 1050 while no Runway model appears in the rows retrieved (live page, no date)."}]}}}}},"microsoft-ai":{"a":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"microsoft-ai","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-23","functions":{"speech-voice":{"alternatives":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/speech-to-text","date":"undated","support":"Microsoft AI's MAI-Transcribe-2 ranks 2nd (2.0% word error), but Alibaba, Mistral, ElevenLabs, xAI, StepFun and Google are within about 0.6 points. Page shows no publication date (date form: undated); read 2026-10-23."},{"url":"https://artificialanalysis.ai/text-to-speech/leaderboard","date":"undated","support":"No Microsoft voice is in the TTS arena top 15, which ElevenLabs, Alibaba, Cartesia, Google and Inworld lead. Page shows no publication date (date form: undated); read 2026-10-23."}]},"switchingTime":{"level":1,"evidence":[{"url":"https://livekit.com/blog/introducing-livekit-inference","date":"2025-10-01","support":"LiveKit says swapping a voice in an agent pipeline is as easy as changing the TTS string; STT and TTS from several vendors behind one key."},{"url":"https://docs.livekit.io/agents/integrations/","date":"undated","support":"LiveKit docs: STT, TTS and realtime models are AgentSession arguments; OpenAI-compatible endpoints can be swapped by base URL. Page shows no publication date (date form: undated); read 2026-10-23."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/speech-to-text","date":"undated","support":"Alibaba's Fun-ASR is more accurate (1.7%) than MAI-Transcribe-2 (2.0%), with Mistral and ElevenLabs close; ElevenLabs and others lead TTS: no gap. Page shows no publication date (date form: undated); read 2026-10-23."}]}}}},"b":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"microsoft-ai","rater":"b","model":"claude-sonnet-5-5","ratedAt":"2026-10-23","functions":{"speech-voice":{"alternatives":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/speech-to-text","date":"undated","support":"MAI-Transcribe-2 scores 2.0% AA-WER (rank 2); Alibaba 1.7%, Mistral 2.1%, ElevenLabs 2.2%, xAI 2.3% and Google 2.6% are close (live page, no date)."},{"url":"https://artificialanalysis.ai/text-to-speech/leaderboard","date":"undated","support":"MAI-Voice-1 scores 1030 (rank 76) and Azure HD 2.5 1142 (rank 30) on TTS (live page, no date)."}]},"switchingTime":{"level":1,"evidence":[{"url":"https://microsoft.ai/news/today-were-announcing-3-new-world-class-mai-models-available-in-foundry/","date":"2026-04-02","support":"MAI-Voice-1 ($22 per million characters) and MAI-Transcribe-1 ($0.36 per hour) are sold through Foundry as standard API models (2 Apr 2026); evidence on enterprise migration time is thin."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/text-to-speech/leaderboard","date":"undated","support":"On TTS, 70+ models outscore MAI-Voice-1; on STT Alibaba is 0.3 points of AA-WER ahead, so the best alternative is not behind (live pages, no date)."}]}}}}},"soundhound":{"a":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"soundhound","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-23","functions":{"speech-voice":{"alternatives":{"level":1,"evidence":[{"url":"https://www.automotiveworld.com/analysis/soundhound-oems-embrace-ai-powered-software-centric-ux/","date":"2024-09-24","support":"In cars, Cerence's Chat Pro reached VW Group models with very similar capabilities aimed at identical use cases"},{"url":"https://artificialanalysis.ai/speech-to-speech/arena","date":"undated","support":"Voice-agent alternatives from Google, xAI, OpenAI, Cartesia and ElevenLabs rank on the arena; SoundHound's enterprise and restaurant voice agents face many rivals. Page shows no publication date (date form: undated); read 2026-10-23."}]},"switchingTime":{"level":2,"evidence":[{"url":"https://www.automotiveworld.com/analysis/soundhound-oems-embrace-ai-powered-software-centric-ux/","date":"2024-09-24","support":"Chat AI is built into Stellantis DS, Peugeot, Opel and Vauxhall cars in 18 countries and 13 languages, delivered over the air; replacing it needs OEM software work (no cycle length stated, thin evidence)"},{"url":"https://livekit.com/blog/introducing-livekit-inference","date":"2025-10-01","support":"LiveKit says swapping a voice in an agent pipeline is as easy as changing the TTS string; STT and TTS from several vendors behind one key."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/speech-to-text","date":"undated","support":"SoundHound does not appear among the leading speech-recognition models; Microsoft, Google, Alibaba and others lead: no capability gap. Page shows no publication date (date form: undated); read 2026-10-23."},{"url":"https://www.automotiveworld.com/analysis/soundhound-oems-embrace-ai-powered-software-centric-ux/","date":"2024-09-24","support":"Cerence offers very similar in-car generative voice capabilities at scale"}]}}}},"b":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"soundhound","rater":"b","model":"claude-sonnet-5-5","ratedAt":"2026-10-23","functions":{"speech-voice":{"alternatives":{"level":3,"evidence":[{"url":"https://www.sec.gov/Archives/edgar/data/1768267/000162828025053372/crnc-20250930.htm","date":"2025-11-20","support":"Cerence's FY2025 10-K names SoundHound and iFlyTek as focused rivals and Amazon, Apple, Google, Microsoft, Alibaba, Baidu and Tencent as assistant providers; rivals have had some success with its customers (filed 2025-11-20)."},{"url":"https://www.sec.gov/Archives/edgar/data/1768267/000162828025053372/crnc-20250930.htm","date":"2025-11-20","support":"For a global carmaker wanting an embedded, brandable assistant, the realistic options are few (Cerence, Google, Amazon, SoundHound); that count is a judgement from the same filing."}]},"switchingTime":{"level":3,"evidence":[{"url":"https://www.sec.gov/Archives/edgar/data/1768267/000162828025053372/crnc-20250930.htm","date":"2025-11-20","support":"Per the filing, development takes roughly six months to two years after a design win, and a carmaker is very unlikely to change the technology until a vehicle model is revamped (Item 1A)."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://www.sec.gov/Archives/edgar/data/1768267/000162828025053372/crnc-20250930.htm","date":"2025-11-20","support":"Cerence cites its scale, global team and broadest language coverage (company claim); no source shows the best alternative falling short of SoundHound, so the gap is rated none."}]}}}}},"iflytek":{"a":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"iflytek","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-23","functions":{"speech-voice":{"alternatives":{"level":2,"evidence":[{"url":"https://artificialanalysis.ai/speech-to-text","date":"undated","support":"Chinese rivals rank at the top of speech recognition: Alibaba Fun-ASR is first (1.7% word error) and StepFun eighth; Alibaba Qwen TTS is third in the TTS arena. Page shows no publication date (date form: undated); read 2026-10-23."},{"url":"https://yicaiglobal.com/news/bytedances-volcano-engine-starts-price-war-in-chinas-enterprise-llms-sector","date":"2024-05-16","support":"ByteDance's Volcano Engine launched speech recognition and speech synthesis models alongside Doubao in May 2024"},{"url":"https://kr-asia.com/iflytek-is-seeing-to-raise-566-5m-to-sharpen-competitive-edge-in-voice-recognition","date":"2018-05-15","support":"Old figure: iFlytek held about 70% of China's voice recognition market, with Alibaba, Tencent and Baidu entering (2018)"}]},"switchingTime":{"level":2,"evidence":[{"url":"https://static.cninfo.com.cn/finalpage/2026-04-29/1225233581.PDF","date":"2026-04-29","support":"The company says its speech scoring serves exams in nearly 30 provinces and 100% of gaokao listening-speaking scoring regions (p.14); public exam and government buyers switch through procurement cycles of months"}]},"capabilityGap":{"level":2,"evidence":[{"url":"https://static.cninfo.com.cn/finalpage/2026-04-29/1225233581.PDF","date":"2026-04-29","support":"The company says it has near-100% share of middle-school English oral exam scoring (p.14), a validated niche where rivals lack deployments"},{"url":"https://artificialanalysis.ai/speech-to-text","date":"undated","support":"In general speech recognition, Alibaba and other developers match or beat iFlytek, so the gap is limited to niche deployments: small. Page shows no publication date (date form: undated); read 2026-10-23."}]}},"avatar-video":{"alternatives":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/video/leaderboard/text-to-video","date":"undated","support":"Chinese developers Alibaba, ByteDance and MiniMax rank near the top of video generation, alongside HeyGen; many can supply presenter video. Page shows no publication date (date form: undated); read 2026-10-23."},{"url":"https://trendingtopics.eu/mirage-raises-75m-to-push-ai-video-app-captions-into-asian-markets/","date":"2026-03-25","support":"Avatar-video competitors named include CapCut (ByteDance), D-ID, HeyGen and Canva"}]},"switchingTime":{"level":2,"evidence":[{"url":"https://static.cninfo.com.cn/finalpage/2026-04-29/1225233581.PDF","date":"2026-04-29","support":"The company says its virtual-human platform serves state-owned, finance and tourism clients, including interactive kiosk hardware, with 100,000+ enterprise digital doubles (p.12, p.41); hardware-linked deployments take months to replace (thin)"}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/video/leaderboard/text-to-video","date":"undated","support":"Several developers lead video generation by a wide margin over models not on the board; no sign iFlytek's avatar models are ahead: no gap (thin evidence). Page shows no publication date (date form: undated); read 2026-10-23."}]}},"robotics-science":{"alternatives":{"level":2,"evidence":[{"url":"https://static.cninfo.com.cn/finalpage/2026-04-29/1225233581.PDF","date":"2026-04-29","support":"The company says over 90% of China's intelligent-robot makers use its robot super-brain platform (Unitree, AgiBot, Galbot) (p.12)"},{"url":"https://artificialanalysis.ai/speech-to-text","date":"undated","support":"Other Chinese voice and model providers (Alibaba, StepFun) rank at the top of speech benchmarks and could supply robot interaction. Page shows no publication date (date form: undated); read 2026-10-23."}]},"switchingTime":{"level":1,"evidence":[{"url":"https://static.cninfo.com.cn/finalpage/2026-04-29/1225233581.PDF","date":"2026-04-29","support":"The company says its AI voice backpack module needs about 2 hours on average for full integration and debugging (p.12): a modular, quick-to-swap component"}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://static.cninfo.com.cn/finalpage/2026-04-29/1225233581.PDF","date":"2026-04-29","support":"The platform is mainly voice and multimodal interaction for robots (p.12), where Chinese rivals' speech models are at least as accurate (see AA speech-to-text): no gap"}]}}}},"b":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"iflytek","rater":"b","model":"claude-sonnet-5-5","ratedAt":"2026-10-23","functions":{"speech-voice":{"alternatives":{"level":2,"evidence":[{"url":"https://www.sec.gov/Archives/edgar/data/1768267/000162828025053372/crnc-20250930.htm","date":"2025-11-20","support":"Cerence's FY2025 10-K names iFlyTek as its rival in China, with Alibaba, Baidu and Tencent also offering assistants (filed 2025-11-20)."},{"url":"https://artificialanalysis.ai/text-to-speech/leaderboard","date":"undated","support":"Chinese vendors in the live English-only arena: Alibaba Qwen-Audio 1301 and 1265, StepFun 1216, Ant Group Ming 1190 (page undated); Mandarin and dialect quality are not covered."}]},"switchingTime":{"level":2,"evidence":[{"url":"https://static.cninfo.com.cn/finalpage/2026-04-29/1225233581.PDF","date":"2026-04-29","support":"2025 annual report p. 12: its AI platform has over 10 million developers and powers 2.4 billion smart devices; p. 31: the company says it holds the top share in in-car cockpits."},{"url":"https://www.sec.gov/Archives/edgar/data/1768267/000162828025053372/crnc-20250930.htm","date":"2025-11-20","support":"Carmakers rarely change voice technology before a model revamp (Cerence 10-K, filed 2025-11-20), which also applies to iFlytek's cockpit and device business."}]},"capabilityGap":{"level":2,"evidence":[{"url":"https://static.cninfo.com.cn/finalpage/2026-04-29/1225233581.PDF","date":"2026-04-29","support":"p. 31: the company says IDC ranked it first on all eight dimensions of its China AI translation evaluation 2025, so rivals fall short somewhat (company-reported third-party result)."},{"url":"https://artificialanalysis.ai/speech-to-text","date":"undated","support":"On the English-only live STT board, Alibaba (1.7% AA-WER) is ahead of most others, which tempers the lead (page undated)."}]}},"avatar-video":{"alternatives":{"level":1,"evidence":[{"url":"https://static.cninfo.com.cn/finalpage/2026-04-29/1225233581.PDF","date":"2026-04-29","support":"p. 12: the virtual-human platform claims 100,000+ enterprise digital twins; p. 41 lists the platform at the development stage in its plan table (2025 annual report)."}]},"switchingTime":{"level":1,"evidence":[{"url":"https://static.cninfo.com.cn/finalpage/2026-04-29/1225233581.PDF","date":"2026-04-29","support":"p. 41: the platform's stated aim is to supply low-cost virtual-human capability to developers, a software tool rather than a deeply embedded system; evidence on switching is thin."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://www.heygen.com/blog/heygen-surpasses-200m-arr","date":"2026-06-25","support":"HeyGen says 85% of the Fortune 100 use its avatar and voice models (25 Jun 2026), so stronger alternatives exist and no shortfall in them is shown."}]}},"robotics-science":{"alternatives":{"level":1,"evidence":[{"url":"https://static.cninfo.com.cn/finalpage/2026-04-29/1225233581.PDF","date":"2026-04-29","support":"p. 12: iFlytek's robot 'super-brain' is mainly a voice and interaction layer, which many vendors supply (company says over 90% of domestic robot makers use its platform)."}]},"switchingTime":{"level":1,"evidence":[{"url":"https://static.cninfo.com.cn/finalpage/2026-04-29/1225233581.PDF","date":"2026-04-29","support":"p. 12: the company says its AI voice backpack module takes on average only two hours of full-chain integration and debugging (company claim)."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/text-to-speech/leaderboard","date":"undated","support":"For the voice layer, Alibaba, Cartesia and Google score above many others in the live arena, so no large shortfall in alternatives (page undated)."}]}}}}},"amazon-agi":{"a":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"amazon-agi","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-23","functions":{"speech-voice":{"alternatives":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/speech-to-text","date":"undated","support":"Amazon Nova 2 Pro (4.9% word error) is behind Alibaba, Microsoft AI, Mistral, ElevenLabs, xAI, Google, AssemblyAI and others. Page shows no publication date (date form: undated); read 2026-10-23."},{"url":"https://artificialanalysis.ai/text-to-speech/leaderboard","date":"undated","support":"Amazon Polly is not in the TTS arena top 15, which ElevenLabs, Alibaba, Cartesia, Google and Inworld lead. Page shows no publication date (date form: undated); read 2026-10-23."},{"url":"https://artificialanalysis.ai/speech-to-speech/arena","date":"undated","support":"Amazon Nova 2.0 Sonic ranks 14th in the voice-agent arena, behind Google, xAI, OpenAI, Cartesia and ElevenLabs. Page shows no publication date (date form: undated); read 2026-10-23."}]},"switchingTime":{"level":1,"evidence":[{"url":"https://livekit.com/blog/introducing-livekit-inference","date":"2025-10-01","support":"LiveKit says swapping a voice in an agent pipeline is as easy as changing the TTS string; STT and TTS from several vendors behind one key."},{"url":"https://docs.livekit.io/agents/integrations/","date":"undated","support":"LiveKit docs: STT, TTS and realtime models are AgentSession arguments; OpenAI-compatible endpoints can be swapped by base URL. Page shows no publication date (date form: undated); read 2026-10-23."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/speech-to-text","date":"undated","support":"Amazon's best transcription model (Nova 2 Pro, 4.9% word error) trails a dozen-plus rivals; no Amazon model is in the TTS top 15: no gap. Page shows no publication date (date form: undated); read 2026-10-23."}]}}}},"b":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"amazon-agi","rater":"b","model":"claude-sonnet-5-5","ratedAt":"2026-10-23","functions":{"speech-voice":{"alternatives":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/text-to-speech/leaderboard","date":"undated","support":"Polly Generative scores 1066 (rank 56) and Polly Long-Form 1029; more than 50 models from other vendors score higher (live page, no date)."},{"url":"https://artificialanalysis.ai/speech-to-text","date":"undated","support":"Amazon Transcribe scores 4.1% AA-WER against 1.7% for the leader (live page, no date)."}]},"switchingTime":{"level":1,"evidence":[{"url":"https://developers.deepgram.com/docs/migrating-from-amazon-web-services-aws-transcribe-to-deepgram","date":"undated","support":"Deepgram's guide shows moving off AWS Transcribe as test audio, response parsing, traffic swap and rollback with no stated timeline (undated); evidence for enterprise contact-centre integration is thin."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/text-to-speech/leaderboard","date":"undated","support":"Alternatives lead by 200+ Elo on TTS (1335 against Polly's 1066) and by 2+ points of AA-WER on STT (live pages, no date)."}]}}}}},"black-forest-labs":{"a":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"black-forest-labs","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-23","functions":{"image-video":{"alternatives":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/image/leaderboard/text-to-image","date":"undated","support":"FLUX 3 ranks 13th in text-to-image behind OpenAI, Google, Microsoft AI, xAI and Meta. Page shows no publication date (date form: undated); read 2026-10-23."},{"url":"https://www.adobe.com/products/firefly/partner-models.html","date":"undated","support":"Adobe lists FLUX next to Google Nano Banana and OpenAI GPT Image as partner image models. Page shows no publication date (date form: undated); read 2026-10-23."}]},"switchingTime":{"level":1,"evidence":[{"url":"https://www.adobe.com/products/firefly/partner-models.html","date":"undated","support":"Adobe's Firefly app offers Google, OpenAI, Black Forest Labs, Runway, Kling and Luma models side by side: switching is a menu choice. Page shows no publication date (date form: undated); read 2026-10-23."},{"url":"https://vercel.com/blog/ai-gateway-production-index-september-2026","date":"2026-09-17","support":"Vercel's gateway shows image and video share moving within weeks (Grok Imagine 42% to 19% of videos generated, June to August; Veo 15% to 20% of video spend in a month)"}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/video/leaderboard/text-to-video","date":"undated","support":"FLUX 3 video ranks 6th and 8th, with Alibaba, Utopai, HeyGen, MiniMax and ByteDance above: no gap. Page shows no publication date (date form: undated); read 2026-10-23."}]}}}},"b":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"black-forest-labs","rater":"b","model":"claude-sonnet-5-5","ratedAt":"2026-10-23","functions":{"image-video":{"alternatives":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/image/leaderboard/text-to-image","date":"undated","support":"FLUX 3 Image scores 1117 (rank 13); OpenAI (1222), Google (1176), Microsoft (1166), xAI (1166) and Meta (1127) score higher (live page, no date)."},{"url":"https://artificialanalysis.ai/video/leaderboard/text-to-video","date":"undated","support":"FLUX 3 scores 1145 (rank 8) on video, near Wan 3.0, Utopai, HeyGen, MiniMax and Seedance (live page, no date)."}]},"switchingTime":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/video/leaderboard/text-to-video","date":"undated","support":"The arena lists FLUX 3 Max and MiniMax H3 Max both served through Fal, a model-hosting platform where many suppliers sit side by side (live page, no date)."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/image/leaderboard/text-to-image","date":"undated","support":"GPT Image 2.5 scores 1222 against FLUX 3 Image's 1117, so the best alternative is ahead (live page, no date)."}]}}}}},"field-ai":{"a":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"field-ai","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-23","functions":{"robotics-science":{"alternatives":{"level":2,"evidence":[{"url":"https://thenextweb.com/news/fieldai-10b-europe-robot-brains","date":"2026-10-02","support":"TNW names Skild AI, Physical Intelligence, Genesis AI and Agile Robots as rivals building general robot brains"},{"url":"https://www.skild.ai/blogs/skild-crosses-100m-arr","date":"2026-09-10","support":"Skild says mobility (inspection, delivery, AMRs) is 10% of its $100m ARR: a direct alternative for site robots"}]},"switchingTime":{"level":2,"evidence":[{"url":"https://www.skild.ai/blogs/skild-crosses-100m-arr","date":"2026-09-10","support":"Robot-brain deployments involve installation, integration and re-training for each site change, so replacing one takes months (evidence from a rival's account)"},{"url":"https://thenextweb.com/news/fieldai-10b-europe-robot-brains","date":"2026-10-02","support":"FieldAI serves construction, data-centre and defence customers (30+); no switching times reported (thin)"}]},"capabilityGap":{"level":2,"evidence":[{"url":"https://thenextweb.com/news/fieldai-10b-europe-robot-brains","date":"2026-10-02","support":"FieldAI reports revenue and contracts past $135m with 30+ customers, more than rivals' mobility revenue; Skild's mobility line is smaller: small gap"},{"url":"https://www.skild.ai/blogs/skild-crosses-100m-arr","date":"2026-09-10","support":"The company says mobility is 10% of its revenue; on $100m+ ARR that is roughly $10m (5LayerCake derivation)"}]}}}},"b":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"field-ai","rater":"b","model":"claude-sonnet-5-5","ratedAt":"2026-10-23","functions":{"robotics-science":{"alternatives":{"level":3,"evidence":[{"url":"https://thenextweb.com/news/fieldai-10b-europe-robot-brains","date":"2026-10-02","support":"Comparables named: Skild AI ($100M run-rate), Physical Intelligence, NEURA, Humanoid, Genesis AI and Agile Robots (2 Oct 2026)."},{"url":"https://www.robotics247.com/article/nvidia-gtc-2026-nvidia-global-robotics-leaders-look-to-take-physical-ai-to-the-real-world","date":"2026-03-19","support":"NVIDIA's open GR00T and Google's Gemini Robotics are other brain options for site robots (19 Mar 2026)."}]},"switchingTime":{"level":2,"evidence":[{"url":"https://siliconangle.com/2025/08/20/robotics-software-startup-fieldai-raises-405m-funding","date":"2025-08-20","support":"FieldAI says one software stack runs on humanoids, vehicles and other systems without custom work per type, deployed on hundreds of sites (20 Aug 2025); re-validating risk limits at each site is judged to take months."}]},"capabilityGap":{"level":2,"evidence":[{"url":"https://thenextweb.com/news/fieldai-10b-europe-robot-brains","date":"2026-10-02","support":"A person familiar says FieldAI's revenue and contracts passed $135M across 30+ customers; Skild has $100M run-rate (2 Oct 2026)."},{"url":"https://www.skild.ai/blogs/skild-crosses-100m-arr","date":"2026-09-10","support":"Skild says about 90% of its revenue is manipulation and 10% mobility (10 Sep 2026), so rivals have less proven site-autonomy revenue."}]}}}}},"deepgram":{"a":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"deepgram","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-23","functions":{"speech-voice":{"alternatives":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/speech-to-text","date":"undated","support":"Over a dozen developers beat Deepgram Nova-3 (5.2% word error) on accuracy, e.g. Alibaba 1.7%, Microsoft 2.0%, ElevenLabs 2.2%, Google 2.6%, AssemblyAI 3.0%: many alternatives. Page shows no publication date (date form: undated); read 2026-10-23."},{"url":"https://livekit.com/blog/introducing-livekit-inference","date":"2025-10-01","support":"LiveKit Inference launched with STT from AssemblyAI, Cartesia and Deepgram side by side"}]},"switchingTime":{"level":1,"evidence":[{"url":"https://livekit.com/blog/introducing-livekit-inference","date":"2025-10-01","support":"LiveKit says swapping a voice in an agent pipeline is as easy as changing the TTS string; STT and TTS from several vendors behind one key."},{"url":"https://docs.livekit.io/agents/integrations/","date":"undated","support":"LiveKit docs: STT, TTS and LLM are set as arguments of AgentSession, and models can be changed during a session. Page shows no publication date (date form: undated); read 2026-10-23."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/speech-to-text","date":"undated","support":"Several alternatives are more accurate than Deepgram's best model (Nova-3, 5.2% AA word error rate): no gap. Page shows no publication date (date form: undated); read 2026-10-23."}]}}}},"b":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"deepgram","rater":"b","model":"claude-sonnet-5-5","ratedAt":"2026-10-23","functions":{"speech-voice":{"alternatives":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/speech-to-text","date":"undated","support":"Live STT board (no date): Deepgram Nova-3 scores 5.2% AA-WER; Alibaba 1.7%, Microsoft 2.0%, Mistral 2.1%, ElevenLabs 2.2%, xAI 2.3%, Google 2.6%, AssemblyAI 3.0% all do better."},{"url":"https://docs.pipecat.ai/server/services/supported-services","date":"undated","support":"Pipecat lists about 35 speech-to-text and more than 50 text-to-speech services that can be plugged in (page undated)."}]},"switchingTime":{"level":1,"evidence":[{"url":"https://developers.deepgram.com/docs/migrating-from-amazon-web-services-aws-transcribe-to-deepgram","date":"undated","support":"Deepgram's own guide for moving off AWS Transcribe is a test, parse-change, swap-traffic and rollback process with no stated duration; an API-level move fits weeks (undated)."},{"url":"https://techcrunch.com/2026/01/13/deepgram-raises-130m-at-1-3b-valuation-and-buys-a-yc-ai-startup","date":"2026-01-13","support":"Customers named are developer platforms: Vapi, Twilio and Granola; 1,300+ organisations in total (13 Jan 2026)."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/speech-to-text","date":"undated","support":"Six or more vendors beat Nova-3's 5.2% AA-WER on the live board (no date), so the best alternative is ahead, not behind."},{"url":"https://artificialanalysis.ai/text-to-speech/leaderboard","date":"undated","support":"Deepgram's Flux TTS scores 1052 (rank 65) against 1335 for the leader (live page, no date)."}]}}}}},"topaz-labs":{"a":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"topaz-labs","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-23","functions":{"image-video":{"alternatives":{"level":2,"evidence":[{"url":"https://thenextweb.com/news/adobe-acquires-topaz-labs-ai-enhancement","date":"2026-06-25","support":"TNW names Freepik, now Magnific, as a rival with a profitable upscaling business, and says the deal takes a strong enhancement player off the board"},{"url":"https://news.adobe.com/news/2026/06/adobe-to-acquire-topaz-labs","date":"2026-06-25","support":"Adobe plans to fold Topaz models into Firefly and Creative Cloud; with Magnific, the evidence names few strong enhancement suppliers (thin), rated several"}]},"switchingTime":{"level":1,"evidence":[{"url":"https://news.adobe.com/news/2026/06/adobe-to-acquire-topaz-labs","date":"2026-06-25","support":"Adobe says Topaz products will remain available standalone; enhancement is a post-processing step that tools can replace within weeks (switching evidence thin)"}]},"capabilityGap":{"level":2,"evidence":[{"url":"https://news.adobe.com/news/2026/06/adobe-to-acquire-topaz-labs","date":"2026-06-25","support":"Adobe calls Topaz's enhancement models industry-leading (acquirer's claim); used by 20 of the 50 largest companies"},{"url":"https://thenextweb.com/news/adobe-acquires-topaz-labs-ai-enhancement","date":"2026-06-25","support":"TNW says Topaz products are widely trusted among professionals and won a 2025 Emmy: a small quality edge"}]}}}},"b":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"topaz-labs","rater":"b","model":"claude-sonnet-5-5","ratedAt":"2026-10-23","functions":{"image-video":{"alternatives":{"level":2,"evidence":[{"url":"https://invideo.io/faq/what-is-the-best-ai-video-upscaler-for-post-production/","date":"2026-07-27","support":"A vendor FAQ (last updated 27 Jul 2026) names four other upscalers: DaVinci Resolve Super Scale, Imagera, Video2X and ViZoUpscale, and calls Topaz the quality benchmark; none is shown at parity."}]},"switchingTime":{"level":1,"evidence":[{"url":"https://news.adobe.com/news/2026/06/adobe-to-acquire-topaz-labs","date":"2026-06-25","support":"Adobe says Topaz tools stay on sale as stand-alone software used by millions, including 20 of the world's 50 largest companies (25 Jun 2026); a file-level tool is swapped in days (judgement)."}]},"capabilityGap":{"level":2,"evidence":[{"url":"https://invideo.io/faq/what-is-the-best-ai-video-upscaler-for-post-production/","date":"2026-07-27","support":"The FAQ says Topaz is highest quality but slowest, with Resolve Super Scale and Imagera below it (vendor source, last updated 27 Jul 2026)."},{"url":"https://news.adobe.com/news/2026/06/adobe-to-acquire-topaz-labs","date":"2026-06-25","support":"Adobe is buying Topaz to build its enhancement models into Firefly and Creative Cloud (25 Jun 2026), a sign of a quality lead."}]}}}}},"assemblyai":{"a":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"assemblyai","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-23","functions":{"speech-voice":{"alternatives":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/speech-to-text","date":"undated","support":"AssemblyAI Universal-3.5 Pro ranks 15th (3.0% word error); Alibaba, Microsoft AI, Mistral, ElevenLabs, xAI, StepFun and Google rank above it: many alternatives. Page shows no publication date (date form: undated); read 2026-10-23."},{"url":"https://livekit.com/blog/introducing-livekit-inference","date":"2025-10-01","support":"LiveKit Inference offers STT from AssemblyAI, Cartesia and Deepgram interchangeably"}]},"switchingTime":{"level":1,"evidence":[{"url":"https://livekit.com/blog/introducing-livekit-inference","date":"2025-10-01","support":"LiveKit says swapping a voice in an agent pipeline is as easy as changing the TTS string; STT and TTS from several vendors behind one key."},{"url":"https://docs.livekit.io/agents/integrations/","date":"undated","support":"LiveKit docs: STT providers are selected per AgentSession argument and can be changed during a session. Page shows no publication date (date form: undated); read 2026-10-23."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/speech-to-text","date":"undated","support":"At least 14 models beat AssemblyAI's best on the AA word-error benchmark: no capability gap. Page shows no publication date (date form: undated); read 2026-10-23."}]}}}},"b":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"assemblyai","rater":"b","model":"claude-sonnet-5-5","ratedAt":"2026-10-23","functions":{"speech-voice":{"alternatives":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/speech-to-text","date":"undated","support":"Live STT board (no date): Universal-3.5 Pro scores 3.0% AA-WER (rank 14); Alibaba 1.7%, Microsoft 2.0%, Mistral 2.1%, ElevenLabs 2.2%, xAI 2.3%, Google 2.6% score better."},{"url":"https://docs.pipecat.ai/server/services/supported-services","date":"undated","support":"AssemblyAI is one of about 35 speech-to-text services Pipecat supports (page undated)."}]},"switchingTime":{"level":1,"evidence":[{"url":"https://developers.deepgram.com/docs/migrating-from-amazon-web-services-aws-transcribe-to-deepgram","date":"undated","support":"A rival's guide for moving off AWS Transcribe describes API-level steps (test audio, response parsing, traffic swap, rollback) with no timeline; used here as a proxy for any STT move, evidence is thin (undated)."},{"url":"https://docs.pipecat.ai/server/services/supported-services","date":"undated","support":"Integrators using Pipecat can select among about 35 STT services by service name (page undated)."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/speech-to-text","date":"undated","support":"Six vendors have lower AA-WER than AssemblyAI's 3.0%, with the best at 1.7% (live page, no date)."}]}}}}},"xai":{"a":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"xai","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-23","functions":{"image-video":{"alternatives":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/image/leaderboard/text-to-image","date":"undated","support":"Grok Imagine Image 2.0 ranks 6th; OpenAI, Google and Microsoft AI rank above or level. Page shows no publication date (date form: undated); read 2026-10-23."},{"url":"https://artificialanalysis.ai/video/leaderboard/text-to-video","date":"undated","support":"Grok Imagine Video 1.5 ranks 12th behind Alibaba, MiniMax, ByteDance, BFL and Google. Page shows no publication date (date form: undated); read 2026-10-23."}]},"switchingTime":{"level":1,"evidence":[{"url":"https://vercel.com/blog/ai-gateway-production-index-september-2026","date":"2026-09-17","support":"Grok Imagine fell from 42% of videos generated on Vercel's gateway in June to 19% in August: users moved within weeks"}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/video/leaderboard/text-to-video","date":"undated","support":"More than ten video models score above Grok Imagine: no gap. Page shows no publication date (date form: undated); read 2026-10-23."}]}}}},"b":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"xai","rater":"b","model":"claude-sonnet-5-5","ratedAt":"2026-10-23","functions":{"image-video":{"alternatives":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/image/leaderboard/text-to-image","date":"undated","support":"Grok Imagine Image 2.0 scores 1166 (rank 6); OpenAI, Google and Microsoft match or beat it (live page, no date)."},{"url":"https://artificialanalysis.ai/video/leaderboard/text-to-video","date":"undated","support":"Grok Imagine Video 1.5 scores 1056 (rank 12); ten other vendors' models score higher (live page, no date). xAI is listed as SpaceXAI."}]},"switchingTime":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/video/leaderboard/image-to-video","date":"undated","support":"Grok Imagine Video 1.5 sits at rank 9 (1067) on image-to-video in a tight field of models from eleven or more vendors; switching evidence is thin (live page, no date)."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/image/leaderboard/text-to-image","date":"undated","support":"GPT Image 2.5 (1222) and Nano Banana 2.1 (1176) outscore Grok Imagine 2.0 (1166) on images (live page, no date)."}]}}}}},"meshy":{"a":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"meshy","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-23","functions":{"image-video":{"alternatives":{"level":2,"evidence":[{"url":"https://www.3dnatives.com/en/tripo-ai-raises-3-billion-yuan-04092026/","date":"2026-09-04","support":"Tripo AI raised about 3bn yuan in September 2026 for 3D foundation models, described as a similar platform to Meshy"},{"url":"https://www.prnewswire.com/news-releases/meshy-surpasses-100-million-in-annual-recurring-revenue-growing-100x-in-under-two-years-302894078.html","date":"2026-09-30","support":"The company benchmarks Meshy 7 against several unnamed comparable 3D systems"}]},"switchingTime":{"level":1,"evidence":[{"url":"https://www.prnewswire.com/news-releases/meshy-surpasses-100-million-in-annual-recurring-revenue-growing-100x-in-under-two-years-302894078.html","date":"2026-09-30","support":"The company says its models export to standard formats such as STL and 3MF, so assets and workflows move easily"}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://www.prnewswire.com/news-releases/meshy-surpasses-100-million-in-annual-recurring-revenue-growing-100x-in-under-two-years-302894078.html","date":"2026-09-30","support":"The company says Meshy 7 scores 5.6 points above the average of comparable systems on texture alignment: a modest lead"},{"url":"https://www.3dnatives.com/en/tripo-ai-raises-3-billion-yuan-04092026/","date":"2026-09-04","support":"Tripo previewed P2.0 with quad topology, an industry first, showing rivals at or above parity on some features"}]}}}},"b":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"meshy","rater":"b","model":"claude-sonnet-5-5","ratedAt":"2026-10-23","functions":{"image-video":{"alternatives":{"level":1,"evidence":[{"url":"https://app.cinevva.com/guides/ai-3d-model-generators","date":"2026-07-17","support":"A July 2026 guide (sold by one listed vendor) lists hosted rivals Tripo, Rodin Gen-2.5, Hi3D and Tencent Hunyuan 3D, and open TRELLIS.2; it says to start with Meshy or Tripo (published 17 Jul 2026, updated 14 Sep)."}]},"switchingTime":{"level":1,"evidence":[{"url":"https://www.prnewswire.com/news-releases/meshy-surpasses-100-million-in-annual-recurring-revenue-growing-100x-in-under-two-years-302894078.html","date":"2026-09-30","support":"Meshy says 15 million registered users and 3,000+ companies (30 Sep 2026); no integration lock-in is named, and a 3D asset tool is swapped per project."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://www.prnewswire.com/news-releases/meshy-surpasses-100-million-in-annual-recurring-revenue-growing-100x-in-under-two-years-302894078.html","date":"2026-09-30","support":"Meshy says its model leads benchmarks but names no comparison systems (company claim); the guide above puts Tripo beside Meshy as a starting point, so no independent gap is shown."}]}}}}},"photoroom":{"a":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"photoroom","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-23","functions":{"image-video":{"alternatives":{"level":1,"evidence":[{"url":"https://techcrunch.com/2024/02/27/confirmed-photoroom-the-ai-image-editor-raised-43m-at-a-500m-valuation/","date":"2024-02-27","support":"TechCrunch calls it a pretty competitive market, naming Picsart and Pixelcut as rivals"}]},"switchingTime":{"level":1,"evidence":[{"url":"https://techcrunch.com/2024/02/27/confirmed-photoroom-the-ai-image-editor-raised-43m-at-a-500m-valuation/","date":"2024-02-27","support":"Photoroom is sold as an app, web tool and API; users and API integrators can move to rivals within weeks (thin)"},{"url":"https://vercel.com/blog/ai-gateway-production-index-september-2026","date":"2026-09-17","support":"Vercel's gateway shows image and video share moving within weeks (Grok Imagine 42% to 19% of videos generated, June to August; Veo 15% to 20% of video spend in a month)"}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/image/leaderboard/text-to-image","date":"undated","support":"Many general image models from OpenAI, Google, Microsoft AI and others now edit images at high quality: no gap (thin, no direct product-photo benchmark). Page shows no publication date (date form: undated); read 2026-10-23."}]}}}},"b":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"photoroom","rater":"b","model":"claude-sonnet-5-5","ratedAt":"2026-10-23","functions":{"image-video":{"alternatives":{"level":1,"evidence":[{"url":"https://techcrunch.com/2024/02/27/confirmed-photoroom-the-ai-image-editor-raised-43m-at-a-500m-valuation","date":"2024-02-27","support":"TechCrunch named Picsart and Pixelcut as rivals and said Photoroom processes about 5 billion images a year (27 Feb 2024); the 2024 date is old and thin."},{"url":"https://www.gradually.ai/en/ai-models/midjourney-v8-1/","date":"undated","support":"On a one-image product-photography test, GPT Image 2 ranked 3rd of 30 and FLUX.2 Max 12th, so general models also do product shots (page undated)."}]},"switchingTime":{"level":1,"evidence":[{"url":"https://techcrunch.com/2024/02/27/confirmed-photoroom-the-ai-image-editor-raised-43m-at-a-500m-valuation","date":"2024-02-27","support":"Photoroom is a freemium app plus API with a web interface (27 Feb 2024); no enterprise lock-in is described."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://www.gradually.ai/en/ai-models/midjourney-v8-1/","date":"undated","support":"GPT Image 2 scored 83.9 on the product-photography test, 3rd best of the models tested, so general models can lead the niche (page undated, one image per model)."}]}}}}},"pixverse":{"a":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"pixverse","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-23","functions":{"image-video":{"alternatives":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/video/leaderboard/text-to-video","date":"undated","support":"PixVerse V6 ranks 20th; Alibaba, MiniMax, ByteDance, Google, xAI and Kling rank above it. Page shows no publication date (date form: undated); read 2026-10-23."},{"url":"https://techcrunch.com/2026/07/13/video-generation-startup-pixverse-raises-439m-valuation-soars-past-2b/","date":"2026-07-13","support":"TechCrunch lists Seedance, Video Rebirth, Kling, Midjourney, Runway and Luma as competitors"}]},"switchingTime":{"level":1,"evidence":[{"url":"https://vercel.com/blog/ai-gateway-production-index-september-2026","date":"2026-09-17","support":"Vercel's gateway shows image and video share moving within weeks (Grok Imagine 42% to 19% of videos generated, June to August; Veo 15% to 20% of video spend in a month)"},{"url":"https://techcrunch.com/2026/07/13/video-generation-startup-pixverse-raises-439m-valuation-soars-past-2b/","date":"2026-07-13","support":"The company says it has 150m+ registered users and 15m monthly actives on a consumer product priced per minute: low switching barriers"}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/video/leaderboard/text-to-video","date":"undated","support":"Nineteen video models score above PixVerse V6: no gap. Page shows no publication date (date form: undated); read 2026-10-23."}]}}}},"b":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"pixverse","rater":"b","model":"claude-sonnet-5-5","ratedAt":"2026-10-23","functions":{"image-video":{"alternatives":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/video/leaderboard/text-to-video","date":"undated","support":"PixVerse V6 scores 984 (rank 20) on text-to-video; at least ten vendors' models rank above it (live page, no date)."},{"url":"https://artificialanalysis.ai/video/leaderboard/image-to-video","date":"undated","support":"PixVerse V6 scores 1024 (rank 15) on image-to-video (live page, no date)."}]},"switchingTime":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/video/leaderboard/image-to-video","date":"undated","support":"Fourteen models from nine or more vendors outrank PixVerse V6 on image-to-video, so users can switch with a new subscription (live page, no date)."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/video/leaderboard/text-to-video","date":"undated","support":"Wan 3.0 (1175) is 191 Elo above PixVerse V6 (984) (live page, no date)."}]}}}}},"alibaba-qwen":{"a":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"alibaba-qwen","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-23","functions":{"image-video":{"alternatives":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/video/leaderboard/text-to-video","date":"undated","support":"Wan 3.0 leads text-to-video (Elo 1175) but Utopai (1167), HeyGen (1160), MiniMax and ByteDance (1158) are within 17 points. Page shows no publication date (date form: undated); read 2026-10-23."},{"url":"https://artificialanalysis.ai/image/leaderboard/text-to-image","date":"undated","support":"Qwen-Image 3.0 ranks 17th-18th in text-to-image behind OpenAI, Google, Microsoft AI, xAI and others. Page shows no publication date (date form: undated); read 2026-10-23."}]},"switchingTime":{"level":1,"evidence":[{"url":"https://vercel.com/blog/ai-gateway-production-index-september-2026","date":"2026-09-17","support":"Vercel's gateway shows image and video share moving within weeks (Grok Imagine 42% to 19% of videos generated, June to August; Veo 15% to 20% of video spend in a month)"}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/video/leaderboard/text-to-video","date":"undated","support":"Wan's lead over four rivals is under 20 Elo: no meaningful gap. Page shows no publication date (date form: undated); read 2026-10-23."}]}}}},"b":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"alibaba-qwen","rater":"b","model":"claude-sonnet-5-5","ratedAt":"2026-10-23","functions":{"image-video":{"alternatives":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/video/leaderboard/text-to-video","date":"undated","support":"Wan 3.0 leads the live text-to-video arena at 1175; Utopai X 1167, HeyGen Video 1 1160, MiniMax H3 1158 and Seedance 2.5 1158 are within 20 Elo (page undated)."},{"url":"https://artificialanalysis.ai/image/leaderboard/text-to-image","date":"undated","support":"Qwen-Image-3.0-Pro scores 1095 (rank 17), behind OpenAI, Google, Microsoft, xAI and Meta (page undated)."}]},"switchingTime":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/video/leaderboard/image-to-video","date":"undated","support":"On image-to-video the same field is tight: MiniMax H3 1140, FLUX 3 Max 1138, Seedance 2.5 1138, Gemini Omni 1132, Wan 3.0 1127, so buyers can choose among many suppliers (live page, no date)."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/video/leaderboard/text-to-video","date":"undated","support":"Wan 3.0 is first on text-to-video and the nearest rival is 8 Elo behind, so the best alternative is at parity (live page, no date)."}]}}}}},"minimax":{"a":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"minimax","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-23","functions":{"music":{"alternatives":{"level":1,"evidence":[{"url":"https://techcrunch.com/2026/02/27/ai-music-generator-suno-hits-2-million-paid-subscribers-and-300m-in-annual-recurring-revenue/","date":"2026-02-27","support":"Suno, the market leader, reports 2m paid subscribers; ElevenLabs, Google and Udio also sell music models"},{"url":"https://decrypt.co/358654/google-ai-music-gemini-lyria-3-review-suno-udio","date":"2026-02-19","support":"Google's Lyria 3, Suno and Udio are all consumer music generators: many alternatives"}]},"switchingTime":{"level":1,"evidence":[{"url":"https://www.wsls.com/business/2025/11/02/ai-song-generator-udio-offers-brief-window-for-downloads-after-universal-settlement-upsets-users/","date":"2025-11-02","support":"AP reports paying Udio users left within days of a download ban: music users switch fast"}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://suno.com/blog/introducing-v6","date":"2026-09-09","support":"Suno says its v6 models are faster and higher quality; a larger, licensed alternative exists: no gap"}]}}}},"b":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"minimax","rater":"b","model":"claude-sonnet-5-5","ratedAt":"2026-10-23","functions":{"music":{"alternatives":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/music/leaderboard/vocals","date":"undated","support":"MiniMax Music 3.0 is the arena's 1000 anchor and 2.5+ scores 987; Suno, Mureka, Google and ElevenLabs all score higher (live page, no date)."}]},"switchingTime":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/music/leaderboard/instrumental","date":"undated","support":"Models from five other vendors (Suno, Mureka, Google, ElevenLabs, Stability) score at or above MiniMax on instrumentals (live page, no date); users switch tools freely."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/music/leaderboard/vocals","date":"undated","support":"Suno v6 (1147) is 147 Elo above MiniMax 3.0, so the best alternative is far ahead (live page, no date)."}]}}}}},"stability-ai":{"a":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"stability-ai","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-23","functions":{"image-video":{"alternatives":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/image/leaderboard/text-to-image","date":"undated","support":"No Stability model is in the text-to-image top 20, which OpenAI, Google, Microsoft AI, xAI, Meta, BFL, ByteDance and Alibaba fill. Page shows no publication date (date form: undated); read 2026-10-23."}]},"switchingTime":{"level":1,"evidence":[{"url":"https://vercel.com/blog/ai-gateway-production-index-september-2026","date":"2026-09-17","support":"Vercel's gateway shows image and video share moving within weeks (Grok Imagine 42% to 19% of videos generated, June to August; Veo 15% to 20% of video spend in a month)"},{"url":"https://www.adobe.com/products/firefly/partner-models.html","date":"undated","support":"Adobe's Firefly app offers Google, OpenAI, Black Forest Labs, Runway, Kling and Luma models side by side: switching is a menu choice. Page shows no publication date (date form: undated); read 2026-10-23."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/image/leaderboard/text-to-image","date":"undated","support":"Many developers rank above Stability's models: no capability gap. Page shows no publication date (date form: undated); read 2026-10-23."}]}}}},"b":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"stability-ai","rater":"b","model":"claude-sonnet-5-5","ratedAt":"2026-10-23","functions":{"image-video":{"alternatives":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/image/leaderboard/text-to-image","date":"undated","support":"No Stability model appears in the rows retrieved from the live text-to-image arena (top 25 plus named vendors); OpenAI, Google, Microsoft, xAI, Meta, Black Forest Labs, ByteDance and Alibaba do (page undated)."}]},"switchingTime":{"level":1,"evidence":[{"url":"https://thenextweb.com/news/stability-ai-76m-funding-universal-warner-sony-music-ea","date":"2026-08-25","support":"Stability now describes itself as a maker of AI for music, gaming and entertainment; the article names no image or video product line (25 Aug 2026), so users have long moved to other models."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/image/leaderboard/text-to-image","date":"undated","support":"GPT Image 2.5 scores 1222 and Nano Banana 2.1 1176, ahead of any Stability model that appears (live page, no date)."}]}}}}},"mirage-captions":{"a":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"mirage-captions","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-23","functions":{"avatar-video":{"alternatives":{"level":1,"evidence":[{"url":"https://trendingtopics.eu/mirage-raises-75m-to-push-ai-video-app-captions-into-asian-markets/","date":"2026-03-25","support":"Competitors named include CapCut (ByteDance), Meta's Edits, Canva, D-ID and HeyGen"},{"url":"https://docs.livekit.io/agents/integrations/avatar/","date":"undated","support":"LiveKit lists 16 avatar video providers. Page shows no publication date (date form: undated); read 2026-10-23."}]},"switchingTime":{"level":1,"evidence":[{"url":"https://trendingtopics.eu/mirage-raises-75m-to-push-ai-video-app-captions-into-asian-markets/","date":"2026-03-25","support":"Mirage runs a freemium mobile and web app for small businesses (Appfigures: $28.4m in-app revenue over 365 days): low switching barriers"}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/video/leaderboard/text-to-video","date":"undated","support":"HeyGen Video 1 ranks 3rd in text-to-video and other developers lead; Mirage is not on the board: no gap. Page shows no publication date (date form: undated); read 2026-10-23."}]}}}},"b":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"mirage-captions","rater":"b","model":"claude-sonnet-5-5","ratedAt":"2026-10-23","functions":{"avatar-video":{"alternatives":{"level":1,"evidence":[{"url":"https://trendingtopics.eu/mirage-raises-75m-to-push-ai-video-app-captions-into-asian-markets/","date":"2026-03-25","support":"Rivals named include ByteDance's CapCut, Meta's Edits, Canva, D-ID, HeyGen and Avataar; Captions has 20M users and $28.4M in-app revenue over 365 days (25 Mar 2026)."}]},"switchingTime":{"level":1,"evidence":[{"url":"https://trendingtopics.eu/mirage-raises-75m-to-push-ai-video-app-captions-into-asian-markets/","date":"2026-03-25","support":"Captions is a freemium consumer editing app (free since Jan 2025) with no lock-in described (25 Mar 2026); users switch apps quickly."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://www.heygen.com/blog/heygen-surpasses-200m-arr","date":"2026-06-25","support":"HeyGen and Synthesia sell stronger avatar models to Fortune 100 buyers, so the best alternative is ahead (25 Jun 2026)."}]}}}}},"hume-ai":{"a":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"hume-ai","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-23","functions":{"speech-voice":{"alternatives":{"level":1,"evidence":[{"url":"https://techcrunch.com/2026/01/22/google-reportedly-snags-up-team-behind-ai-voice-startup-hume-ai","date":"2026-01-22","support":"TechCrunch names OpenAI, Meta (Play AI), ElevenLabs and Google's Gemini Live as voice rivals; Google itself took a non-exclusive licence to Hume technology"},{"url":"https://artificialanalysis.ai/speech-to-speech/arena","date":"undated","support":"Google, xAI, OpenAI, Cartesia and ElevenLabs all field ranked voice-agent models: many alternatives. Page shows no publication date (date form: undated); read 2026-10-23."}]},"switchingTime":{"level":1,"evidence":[{"url":"https://livekit.com/blog/introducing-livekit-inference","date":"2025-10-01","support":"LiveKit says swapping a voice in an agent pipeline is as easy as changing the TTS string; STT and TTS from several vendors behind one key."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://techcrunch.com/2026/01/22/google-reportedly-snags-up-team-behind-ai-voice-startup-hume-ai","date":"2026-01-22","support":"Google DeepMind hired Hume's CEO and about seven engineers and got a non-exclusive licence, so Hume's emotional-voice know-how is already available to a far larger rival"}]}}}},"b":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"hume-ai","rater":"b","model":"claude-sonnet-5-5","ratedAt":"2026-10-23","functions":{"speech-voice":{"alternatives":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/text-to-speech/leaderboard","date":"undated","support":"Hume Octave 2 scores 1055 (rank 63) and Octave TTS 1038 (rank 71); dozens of models from many vendors score higher (live page, no date)."}]},"switchingTime":{"level":1,"evidence":[{"url":"https://docs.livekit.io/agents/models/tts/","date":"undated","support":"Hume is one of about 40 TTS plugins in LiveKit, alongside Cartesia, ElevenLabs, Google and OpenAI (page undated)."},{"url":"https://techcrunch.com/2026/01/22/google-reportedly-snags-up-team-behind-ai-voice-startup-hume-ai","date":"2026-01-22","support":"Its CEO and about seven engineers joined Google DeepMind under a licensing deal; the remaining company keeps supplying its technology (22 Jan 2026)."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/text-to-speech/leaderboard","date":"undated","support":"The leader scores 1335 against Hume's 1055, so the best alternative is far ahead (live page, no date)."}]}}}}},"chai-discovery":{"a":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"chai-discovery","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-23","functions":{"robotics-science":{"alternatives":{"level":2,"evidence":[{"url":"https://www.pharmaceutical-technology.com/news/nabla-takeda-protein-therapeutics-design/","date":"2025-10-15","support":"Nabla Bio's JAM platform reports zero-shot de novo antibody binders and signed a second Takeda deal worth over $1bn in milestones"},{"url":"https://techcrunch.com/2026/01/16/from-openais-offices-to-a-deal-with-eli-lilly-how-chai-discovery-became-one-of-the-flashiest-names-in-ai-drug-development/","date":"2026-01-16","support":"Chai's Lilly deal sits among several AI drug-design partnerships; the article notes rival approaches and sceptics (competitors not named, thin)"}]},"switchingTime":{"level":2,"evidence":[{"url":"https://techcrunch.com/2026/01/16/from-openais-offices-to-a-deal-with-eli-lilly-how-chai-discovery-became-one-of-the-flashiest-names-in-ai-drug-development/","date":"2026-01-16","support":"Lilly uses Chai's software to develop medicines; backers expect partnering in 2026 and first trials by end 2027, so a switch would set programmes back months"}]},"capabilityGap":{"level":2,"evidence":[{"url":"https://www.pharmaceutical-technology.com/news/nabla-takeda-protein-therapeutics-design/","date":"2025-10-15","support":"Nabla reports picomolar zero-shot binders and favourable primate data, a comparable capability claim (company claims, not independently validated)"},{"url":"https://techcrunch.com/2026/01/16/from-openais-offices-to-a-deal-with-eli-lilly-how-chai-discovery-became-one-of-the-flashiest-names-in-ai-drug-development/","date":"2026-01-16","support":"Chai says its codebase uses custom architectures; no independent hit-rate comparison is given: small gap"}]}}}},"b":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"chai-discovery","rater":"b","model":"claude-sonnet-5-5","ratedAt":"2026-10-23","functions":{"robotics-science":{"alternatives":{"level":2,"evidence":[{"url":"https://www.biopharmatrend.com/news/nabla-bio-and-takeda-expand-ai-driven-antibody-design-partnership-in-billion-dollar-deal-1416/","date":"2025-10-14","support":"Nabla Bio signed a multi-year Takeda deal worth over $1B in milestones (14 Oct 2025)."},{"url":"https://www.genengnews.com/topics/artificial-intelligence/scratch-that-de-novo-antibody-design-enters-the-ai-drug-discovery-toolbox/","date":"2025-05-20","support":"Generate Biomedicines has a Novartis collaboration; Absci works with AstraZeneca; the Baker lab's RFdiffusion shows open tools (20 May 2025). Four to five credible players."}]},"switchingTime":{"level":2,"evidence":[{"url":"https://intuitionlabs.ai/articles/generative-antibody-design-chai-3-pfizer","date":"2026-06-13","support":"Pfizer licensed Chai-3 with a custom model trained on its proprietary data (13 Jun 2026); research collaborations with custom models take months to replicate (judgement)."},{"url":"https://www.biopharmatrend.com/news/nabla-bio-and-takeda-expand-ai-driven-antibody-design-partnership-in-billion-dollar-deal-1416/","date":"2025-10-14","support":"Nabla's Takeda collaboration builds on a 2022 partnership (14 Oct 2025), showing programmes run for years."}]},"capabilityGap":{"level":2,"evidence":[{"url":"https://intuitionlabs.ai/articles/generative-antibody-design-chai-3-pfizer","date":"2026-06-13","support":"Chai-3 is said to double Chai-2's roughly 16-20% hit rate (company and press, 13 Jun 2026); no independent head-to-head benchmark is cited, so any shortfall in rivals is small."},{"url":"https://www.genengnews.com/topics/artificial-intelligence/scratch-that-de-novo-antibody-design-enters-the-ai-drug-discovery-toolbox/","date":"2025-05-20","support":"Competitors' de novo claims are disputed, for example Absci's HER2 design based on trastuzumab (20 May 2025)."}]}}}}},"leonardo-ai":{},"lightricks":{},"krea":{},"luma-ai":{},"shengshu":{},"cartesia":{"a":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"cartesia","rater":"a","model":"claude-opus-5-5","ratedAt":"2026-10-23","functions":{"speech-voice":{"alternatives":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/text-to-speech/leaderboard","date":"undated","support":"Cartesia Sonic 3.6 ranks 4th (Elo 1289); ElevenLabs, Alibaba, Google, Inworld, Fish Audio and Speechify are close: many alternatives. Page shows no publication date (date form: undated); read 2026-10-23."},{"url":"https://livekit.com/blog/introducing-livekit-inference","date":"2025-10-01","support":"LiveKit Inference launched TTS from Cartesia, ElevenLabs, Inworld and Rime behind one API key"}]},"switchingTime":{"level":1,"evidence":[{"url":"https://livekit.com/blog/introducing-livekit-inference","date":"2025-10-01","support":"LiveKit says swapping a voice in an agent pipeline is as easy as changing the TTS string; STT and TTS from several vendors behind one key."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/text-to-speech/leaderboard","date":"undated","support":"ElevenLabs (1335) and Alibaba (1301) score above Cartesia (1289) in the TTS arena: no gap. Page shows no publication date (date form: undated); read 2026-10-23."}]}}}},"b":{"schema":"5lc-index-rating/1","key":"models-specialist-models","entityId":"cartesia","rater":"b","model":"claude-sonnet-5-5","ratedAt":"2026-10-23","functions":{"speech-voice":{"alternatives":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/text-to-speech/leaderboard","date":"undated","support":"Sonic 3.6 scores 1289 (rank 4); ElevenLabs 1335 and 1329, Alibaba 1301, Google 1278 and Inworld 1267 sit around it (live page, no date)."},{"url":"https://artificialanalysis.ai/articles/fun-realtime-tts-new-text-to-speech-model-topping-artificial-analysis-leaderboard","date":"2026-06-03","support":"Artificial Analysis said just 24 Elo points separated the top five TTS models (Alibaba, Google, Inworld, Cartesia), 3 Jun 2026."}]},"switchingTime":{"level":1,"evidence":[{"url":"https://docs.livekit.io/agents/models/tts/","date":"undated","support":"Cartesia is both a LiveKit Inference model and a plugin among about 40 TTS providers; the page does not say a swap is one line, so evidence is indirect (undated)."},{"url":"https://docs.vapi.ai/providers/voice/elevenlabs","date":"undated","support":"Vapi sets the voice provider with a voice.provider field on the assistant (ElevenLabs example), so switching is a configuration change (docs undated)."}]},"capabilityGap":{"level":1,"evidence":[{"url":"https://artificialanalysis.ai/text-to-speech/leaderboard","date":"undated","support":"ElevenLabs v4 Turbo (1335) and Alibaba (1301) outscore Cartesia's best model (1289), so the best alternative is ahead (live page, no date)."}]}}}}},"decart":{}},"validation":[]}