{ "data": { "checked": "2026-10-23", "released": [], "waiting": [ { "access": "unknown", "announced": "2026-10-23", "announcedHow": "teaser video on Moonshot's @Kimi_Moonshot X account (cited via pages that describe it)", "expected": "2026-10-27", "expectedBy": "a tester on X says before 27 October; Moonshot gave no date", "kind": "language model", "name": "Kimi K3.1", "note": "Successor to Kimi K3; the 24-second teaser shows the Kimi Code terminal running 'Model: Kimi K3.1' and generating Three.js 3D scenes. Believed to be the anonymous 'Kibitz' model on Arena; size, licence and date not stated (a leaker, not Moonshot, says it ships before 27 October).", "overdue": false, "sources": [ { "url": "https://www.testingcatalog.com/kimi-k3-1-spotted-testing-arena/", "what": "TestingCatalog: Moonshot teaser post of 23 Oct and Arena 'Kibitz' sighting (x.com post quoted here, not read directly)", "seen": "2026-10-23" }, { "url": "https://www.kucoin.com/news/flash/kimi-official-video-reveals-k3-1-model", "what": "KuCoin News: describes the official Kimi teaser video and the kimi-k3-1 API identifier", "seen": "2026-10-23" } ], "status": "teased", "vendor": "Moonshot AI" }, { "access": "unknown", "announced": "2026-10-05", "announcedHow": "Reka blog post 'Rho-1: Collapsing the multimodal stack'", "expected": "", "expectedBy": "", "kind": "multimodal model", "name": "Rho-1", "note": "19B omni-reasoning model trained from scratch that reads and generates text, images and video and outputs robot actions in one network. Research preview only: no public API, download or price; collaborators contact Reka.", "overdue": false, "sources": [ { "url": "https://reka.ai/news/rho-1-collapsing-the-multimodal-stack", "what": "Reka announcement, 5 Oct 2026", "seen": "2026-10-23" }, { "url": "https://quasa.io/stories/reka-s-rho-1-unifies-video-and-robot-actions-but-it-is-still-a-preview", "what": "Quasa: access limited to collaborators, no general release date", "seen": "2026-10-23" } ], "status": "preview", "vendor": "Reka" }, { "access": "unknown", "announced": "2026-10-01", "announcedHow": "Tavus research post introducing Griffin", "expected": "", "expectedBy": "", "kind": "video model", "name": "Griffin", "note": "Real-time, full-duplex video-to-video 'Human Interaction Model' for face-to-face AI conversation. A smaller Griffin-Lite is a research preview for select trusted testers (request form); Tavus says the full Griffin follows once safety work is done.", "overdue": false, "sources": [ { "url": "https://www.tavus.io/griffin", "what": "Tavus announcement, 1 Oct 2026: Griffin-Lite preview for trusted testers, Griffin to follow", "seen": "2026-10-23" }, { "url": "https://letsdatascience.com/news/tavus-releases-griffin-live-video-model-4ceaa1bd", "what": "Let's Data Science: no general launch date announced", "seen": "2026-10-23" } ], "status": "preview", "vendor": "Tavus" }, { "access": "unknown", "announced": "2026-09-22", "announcedHow": "Apsara Conference keynote and Alibaba Cloud press release", "expected": "", "expectedBy": "", "kind": "language model", "name": "Qwen 4 (Max, Plus, Flash)", "note": "Next Qwen generation on a new architecture, said to be in training; previewed as Max (flagship), Plus and Flash tiers. No model card, API name, price or benchmarks yet; the Qwen LLM lead said only that it will come 'soon'.", "overdue": false, "sources": [ { "url": "https://www.alibabacloud.com/press-room/alibaba-unveils-roadmap-on-full-stack-ai-strategy", "what": "Alibaba Cloud press release, 22 Sep 2026: Qwen 4 'currently in training'", "seen": "2026-10-23" }, { "url": "https://latenode.com/news/alibaba-previews-qwen-4-lineup-without-api-pricing-or-release-date", "what": "Latenode: lineup previewed at Apsara, no date or pricing", "seen": "2026-10-23" }, { "url": "https://www.datalearner.com/ai-models/pretrained-models/qwen4-plus", "what": "DataLearner: Qwen4-Plus announced, not yet available; Alibaba said 'soon'", "seen": "2026-10-23" } ], "status": "announced", "vendor": "Alibaba (Qwen)" }, { "access": "open weights", "announced": "2026-09-22", "announcedHow": "shown in the Qwen 4 lineup at the Apsara Conference", "expected": "", "expectedBy": "", "kind": "language model", "name": "Qwen 4 27B", "note": "27B model presented as the open, locally deployable tier of the Qwen 4 family. The open-weight plan comes from coverage of the keynote; Alibaba's press release does not spell it out, and no weights are published.", "overdue": false, "sources": [ { "url": "https://www.datalearner.com/ai-models/pretrained-models/qwen4-plus", "what": "DataLearner: lists Qwen4-27B as the open-source tier shown at Apsara", "seen": "2026-10-23" }, { "url": "https://www.orcarouter.ai/blog/qwen-4-max-lineup-announced-apsara-2026", "what": "OrcaRouter: 27B described as the open-weight local tier, no weights yet", "seen": "2026-10-23" } ], "status": "announced", "vendor": "Alibaba (Qwen)" }, { "access": "unknown", "announced": "2026-09-22", "announcedHow": "Alibaba Cloud press release at the Apsara Conference", "expected": "", "expectedBy": "Later in 2026, Alibaba says", "kind": "image model", "name": "Qwen-Image 3.1", "note": "Image generation model following Qwen-Image 3.0. Alibaba said it will launch 'later this year' (2026) but gave no date; no specs published.", "overdue": false, "sources": [ { "url": "https://www.alibabacloud.com/press-room/alibaba-unveils-roadmap-on-full-stack-ai-strategy", "what": "Alibaba Cloud press release: Qwen-Image 3.1 set to launch later this year", "seen": "2026-10-23" }, { "url": "https://www.datalearner.com/ai-models/pretrained-models/qwen-image-3-1", "what": "DataLearner: not yet officially released", "seen": "2026-10-23" } ], "status": "announced", "vendor": "Alibaba (Qwen)" }, { "access": "unknown", "announced": "2026-09-13", "announcedHow": "Elon Musk posts on X (quoted by third-party pages)", "expected": "", "expectedBy": "", "kind": "language model", "name": "Grok 4.8", "note": "Musk described a 2.5-trillion-parameter model on a new C++ training stack that would 'finish training this week and start RL', and later called it a noticeable improvement over Grok 4.7. No model ID, price or date from xAI.", "overdue": false, "sources": [ { "url": "https://www.orcarouter.ai/blog/grok-4-8-leak", "what": "OrcaRouter: quotes Musk's 13-14 Sep X posts (x.com not read directly)", "seen": "2026-10-23" }, { "url": "https://geotoolbox.ai/blog/grok-5", "what": "Geotoolbox: Musk's 14 Sep posts via TeslaNorth, Grok 4.8 and 4.9 before Grok 5", "seen": "2026-10-23" }, { "url": "https://andrew.ooo/answers/grok-4-8-vs-grok-4-7-vs-grok-5-xai-roadmap-october-2026/", "what": "andrew.ooo: still no Grok 4.8 model as of 3 Oct", "seen": "2026-10-23" } ], "status": "teased", "vendor": "xAI" }, { "access": "API", "announced": "2026-09-03", "announcedHow": "press release introducing the OpenEvidence model family", "expected": "", "expectedBy": "", "kind": "language model", "name": "Darwin", "note": "OpenEvidence's most capable medical model, in an application-based research preview for institutional partners, research collaborators and accredited academic AI researchers (partner endpoint). Size not disclosed; capabilities are to move into its production models later.", "overdue": false, "sources": [ { "url": "https://www.01net.it/introducing-the-openevidence-model-family/", "what": "Business Wire release, 3 Sep 2026, republished by 01net", "seen": "2026-10-23" }, { "url": "https://www.iatrox.com/blog/openevidence-darwin-access-explained-research-preview-eligibility-and-the-path-to-wider-use", "what": "iatroX: access by application, no general release date", "seen": "2026-10-23" } ], "status": "preview", "vendor": "OpenEvidence" }, { "access": "unknown", "announced": "2026-08-31", "announcedHow": "named by founder Tang Jie on the half-year results call", "expected": "", "expectedBy": "", "kind": "language model", "name": "GLM-6.0", "note": "Next Z.ai base model, framed around 'self-evolution' (the model judging when to stop and correct its own training). Tang Jie said the base model will grow while active parameters are kept in check; no date or licence given.", "overdue": false, "sources": [ { "url": "https://news.10jqka.com.cn/20260901/c679482901.shtml", "what": "10jqka (Tonghuashun): report of Zhipu's 31 Aug results briefing Q\u0026A", "seen": "2026-10-23" }, { "url": "https://www.orcarouter.ai/blog/glm-6-0-leak", "what": "OrcaRouter: no Z.ai announcement beyond the call", "seen": "2026-10-23" } ], "status": "teased", "vendor": "Zhipu AI (Z.ai)" }, { "access": "unknown", "announced": "2026-07-30", "announcedHow": "roadmap announced at Sarvam's Epoch 2026 conference in Bengaluru", "expected": "2027-01", "expectedBy": "within six months, Sarvam says", "kind": "language model", "name": "Sarvam trillion-parameter model (unnamed)", "note": "Trillion-plus-parameter foundation model trained from scratch in India, aimed at coding, cybersecurity, simulation and science. Sarvam said it expects the model within six months (by about January 2027); no name or licence yet.", "overdue": false, "sources": [ { "url": "https://letsdatascience.com/news/sarvam-ai-plans-trillion-parameter-foundation-model-002360b8", "what": "Let's Data Science: announced 30 Jul 2026 at Epoch", "seen": "2026-10-23" }, { "url": "https://freepressjournal.in/tech/sarvam-bets-big-on-indias-ai-future-with-trillion-parameter-model-will-be-priced-five-times-cheaper-than-global-rivals", "what": "Free Press Journal: cofounder Pratyush Kumar, 'within six months'", "seen": "2026-10-23" } ], "status": "announced", "vendor": "Sarvam AI" }, { "access": "open weights", "announced": "2026-07-23", "announcedHow": "Black Forest Labs FLUX 3 launch post (launch plan section)", "expected": "", "expectedBy": "", "kind": "multimodal model", "name": "FLUX 3 Dev", "note": "Planned open-weight release of the FLUX 3 multimodal backbone for video, audio and image generation and action prediction. The other FLUX 3 parts (Video, Image, Image Klein, Action, Max) have shipped; no date, size or licence for Dev.", "overdue": false, "sources": [ { "url": "https://bfl.ai/blog/flux-3", "what": "BFL FLUX 3 post, 23 Jul 2026: lists 'FLUX 3 Dev' open-weight access as a coming capability", "seen": "2026-10-23" }, { "url": "https://www.scriptbyai.com/flux-timeline/", "what": "ScriptByAI FLUX timeline: Dev still upcoming, no confirmed date", "seen": "2026-10-23" } ], "status": "announced", "vendor": "Black Forest Labs" }, { "access": "open weights", "announced": "2026-03-16", "announcedHow": "NVIDIA press release launching the Nemotron Coalition (GTC)", "expected": "", "expectedBy": "As early as late autumn 2026, The Information reported", "kind": "language model", "name": "Nemotron 4", "note": "Family of open models to be built on the Nemotron Coalition's first model, co-developed with Mistral AI. The Information reported the largest model exceeds 1 trillion parameters and could be ready as early as late fall 2026; NVIDIA has not confirmed a date. Announced slightly more than six months ago.", "overdue": false, "sources": [ { "url": "https://investor.nvidia.com/news/press-release-details/2026/NVIDIA-Launches-Nemotron-Coalition-of-Leading-Global-AI-Labs-to-Advance-Open-Frontier-Models/default.aspx", "what": "NVIDIA press release, 16 Mar 2026: coalition model will underpin the upcoming Nemotron 4 family of open models", "seen": "2026-10-23" }, { "url": "https://www.vktr.com/ai-news/nvidia-developing-1-trillion-parameter-open-model-to-rival-top-ai-systems/", "what": "VKTR, 11 Aug 2026: The Information report on size and late-fall timing", "seen": "2026-10-23" } ], "status": "announced", "vendor": "NVIDIA" }, { "access": "unknown", "announced": "2026-01-06", "announcedHow": "xAI Series E funding post ('Grok 5 is currently in training')", "expected": "", "expectedBy": "", "kind": "language model", "name": "Grok 5", "note": "xAI's next flagship, in training on Colossus 2. Musk has given shifting targets (most recently that it comes after Grok 4.8 and 4.9); no xAI date. Announced more than six months ago but still unreleased.", "overdue": false, "sources": [ { "url": "https://x.ai/blog/series-e", "what": "xAI Series E post, 6 Jan 2026", "seen": "2026-10-23" }, { "url": "https://geotoolbox.ai/blog/grok-5", "what": "Geotoolbox: status and Musk statements to Sep 2026, no official date", "seen": "2026-10-23" } ], "status": "announced", "vendor": "xAI" } ] }, "kind": "anchor.page", "links": { "api": "https://www.anchorterminal.com/api/v1/index.json", "html": "https://www.anchorterminal.com/upcoming/", "json": "https://www.anchorterminal.com/upcoming/index.json", "llms": "https://www.anchorterminal.com/llms.txt", "markdown": "https://www.anchorterminal.com/upcoming/index.md", "mcp": "https://www.anchorterminal.com/mcp", "slim": "https://www.anchorterminal.com/upcoming/index.min.md" }, "markdown": "Models a lab has announced, teased or shown in a closed preview that nobody can download or call yet, with the day each was announced and when it's expected. We list a model when it comes out, so this is what we're watching. 13 waiting.\n\nJSON: https://www.anchorterminal.com/api/v1/upcoming.json\n\n## Not out yet\n\n| Model | Who | Announced | Expected | Status | Access | What's known | Sources |\n| --- | --- | --- | --- | --- | --- | --- | --- |\n| Kimi K3.1 | Moonshot AI | 2026-10-23 (teaser video on Moonshot's @Kimi_Moonshot X account (cited via pages that describe it)) | 27 October 2026 (a tester on X says before 27 October; Moonshot gave no date) | Teased | unknown | Successor to Kimi K3; the 24-second teaser shows the Kimi Code terminal running 'Model: Kimi K3.1' and generating Three.js 3D scenes. Believed to be the anonymous 'Kibitz' model on Arena; size, licence and date not stated (a leaker, not Moonshot, says it ships before 27 October). | \u003chttps://www.testingcatalog.com/kimi-k3-1-spotted-testing-arena/\u003e \u003chttps://www.kucoin.com/news/flash/kimi-official-video-reveals-k3-1-model\u003e |\n| Rho-1 | Reka | 2026-10-05 (Reka blog post 'Rho-1: Collapsing the multimodal stack') | No date given | Closed preview | unknown | 19B omni-reasoning model trained from scratch that reads and generates text, images and video and outputs robot actions in one network. Research preview only: no public API, download or price; collaborators contact Reka. | \u003chttps://reka.ai/news/rho-1-collapsing-the-multimodal-stack\u003e \u003chttps://quasa.io/stories/reka-s-rho-1-unifies-video-and-robot-actions-but-it-is-still-a-preview\u003e |\n| Griffin | Tavus | 2026-10-01 (Tavus research post introducing Griffin) | No date given | Closed preview | unknown | Real-time, full-duplex video-to-video 'Human Interaction Model' for face-to-face AI conversation. A smaller Griffin-Lite is a research preview for select trusted testers (request form); Tavus says the full Griffin follows once safety work is done. | \u003chttps://www.tavus.io/griffin\u003e \u003chttps://letsdatascience.com/news/tavus-releases-griffin-live-video-model-4ceaa1bd\u003e |\n| Qwen 4 (Max, Plus, Flash) | Alibaba (Qwen) | 2026-09-22 (Apsara Conference keynote and Alibaba Cloud press release) | No date given | Announced | unknown | Next Qwen generation on a new architecture, said to be in training; previewed as Max (flagship), Plus and Flash tiers. No model card, API name, price or benchmarks yet; the Qwen LLM lead said only that it will come 'soon'. | \u003chttps://www.alibabacloud.com/press-room/alibaba-unveils-roadmap-on-full-stack-ai-strategy\u003e \u003chttps://latenode.com/news/alibaba-previews-qwen-4-lineup-without-api-pricing-or-release-date\u003e \u003chttps://www.datalearner.com/ai-models/pretrained-models/qwen4-plus\u003e |\n| Qwen 4 27B | Alibaba (Qwen) | 2026-09-22 (shown in the Qwen 4 lineup at the Apsara Conference) | No date given | Announced | open weights | 27B model presented as the open, locally deployable tier of the Qwen 4 family. The open-weight plan comes from coverage of the keynote; Alibaba's press release does not spell it out, and no weights are published. | \u003chttps://www.datalearner.com/ai-models/pretrained-models/qwen4-plus\u003e \u003chttps://www.orcarouter.ai/blog/qwen-4-max-lineup-announced-apsara-2026\u003e |\n| Qwen-Image 3.1 | Alibaba (Qwen) | 2026-09-22 (Alibaba Cloud press release at the Apsara Conference) | Later in 2026, Alibaba says | Announced | unknown | Image generation model following Qwen-Image 3.0. Alibaba said it will launch 'later this year' (2026) but gave no date; no specs published. | \u003chttps://www.alibabacloud.com/press-room/alibaba-unveils-roadmap-on-full-stack-ai-strategy\u003e \u003chttps://www.datalearner.com/ai-models/pretrained-models/qwen-image-3-1\u003e |\n| Grok 4.8 | xAI | 2026-09-13 (Elon Musk posts on X (quoted by third-party pages)) | No date given | Teased | unknown | Musk described a 2.5-trillion-parameter model on a new C++ training stack that would 'finish training this week and start RL', and later called it a noticeable improvement over Grok 4.7. No model ID, price or date from xAI. | \u003chttps://www.orcarouter.ai/blog/grok-4-8-leak\u003e \u003chttps://geotoolbox.ai/blog/grok-5\u003e \u003chttps://andrew.ooo/answers/grok-4-8-vs-grok-4-7-vs-grok-5-xai-roadmap-october-2026/\u003e |\n| Darwin | OpenEvidence | 2026-09-03 (press release introducing the OpenEvidence model family) | No date given | Closed preview | API | OpenEvidence's most capable medical model, in an application-based research preview for institutional partners, research collaborators and accredited academic AI researchers (partner endpoint). Size not disclosed; capabilities are to move into its production models later. | \u003chttps://www.01net.it/introducing-the-openevidence-model-family/\u003e \u003chttps://www.iatrox.com/blog/openevidence-darwin-access-explained-research-preview-eligibility-and-the-path-to-wider-use\u003e |\n| GLM-6.0 | Zhipu AI (Z.ai) | 2026-08-31 (named by founder Tang Jie on the half-year results call) | No date given | Teased | unknown | Next Z.ai base model, framed around 'self-evolution' (the model judging when to stop and correct its own training). Tang Jie said the base model will grow while active parameters are kept in check; no date or licence given. | \u003chttps://news.10jqka.com.cn/20260901/c679482901.shtml\u003e \u003chttps://www.orcarouter.ai/blog/glm-6-0-leak\u003e |\n| Sarvam trillion-parameter model (unnamed) | Sarvam AI | 2026-07-30 (roadmap announced at Sarvam's Epoch 2026 conference in Bengaluru) | January 2027 (within six months, Sarvam says) | Announced | unknown | Trillion-plus-parameter foundation model trained from scratch in India, aimed at coding, cybersecurity, simulation and science. Sarvam said it expects the model within six months (by about January 2027); no name or licence yet. | \u003chttps://letsdatascience.com/news/sarvam-ai-plans-trillion-parameter-foundation-model-002360b8\u003e \u003chttps://freepressjournal.in/tech/sarvam-bets-big-on-indias-ai-future-with-trillion-parameter-model-will-be-priced-five-times-cheaper-than-global-rivals\u003e |\n| FLUX 3 Dev | Black Forest Labs | 2026-07-23 (Black Forest Labs FLUX 3 launch post (launch plan section)) | No date given | Announced | open weights | Planned open-weight release of the FLUX 3 multimodal backbone for video, audio and image generation and action prediction. The other FLUX 3 parts (Video, Image, Image Klein, Action, Max) have shipped; no date, size or licence for Dev. | \u003chttps://bfl.ai/blog/flux-3\u003e \u003chttps://www.scriptbyai.com/flux-timeline/\u003e |\n| Nemotron 4 | NVIDIA | 2026-03-16 (NVIDIA press release launching the Nemotron Coalition (GTC)) | As early as late autumn 2026, The Information reported | Announced | open weights | Family of open models to be built on the Nemotron Coalition's first model, co-developed with Mistral AI. The Information reported the largest model exceeds 1 trillion parameters and could be ready as early as late fall 2026; NVIDIA has not confirmed a date. Announced slightly more than six months ago. | \u003chttps://investor.nvidia.com/news/press-release-details/2026/NVIDIA-Launches-Nemotron-Coalition-of-Leading-Global-AI-Labs-to-Advance-Open-Frontier-Models/default.aspx\u003e \u003chttps://www.vktr.com/ai-news/nvidia-developing-1-trillion-parameter-open-model-to-rival-top-ai-systems/\u003e |\n| Grok 5 | xAI | 2026-01-06 (xAI Series E funding post ('Grok 5 is currently in training')) | No date given | Announced | unknown | xAI's next flagship, in training on Colossus 2. Musk has given shifting targets (most recently that it comes after Grok 4.8 and 4.9); no xAI date. Announced more than six months ago but still unreleased. | \u003chttps://x.ai/blog/series-e\u003e \u003chttps://geotoolbox.ai/blog/grok-5\u003e |\n\n", "meta": { "attribution": "Anchor Terminal (https://www.anchorterminal.com)", "attributionRequired": true, "credit": "CC BY 4.0, credit Anchor Terminal and link the page or API address you took it from (https://www.anchorterminal.com/terms/#data)", "docs": "https://www.anchorterminal.com/docs/", "generatedAt": "2026-10-24", "license": "CC-BY-4.0", "method": "https://www.anchorterminal.com/benchmark/", "methodology": "0.6", "openapi": "https://www.anchorterminal.com/openapi.json", "preview": false, "run": "2026-10-01", "runLabel": "October 2026 research run", "terms": "https://www.anchorterminal.com/terms/#data" }, "page": { "breadcrumbs": [ { "name": "Home", "url": "https://www.anchorterminal.com/" }, { "name": "Coming models", "url": "" } ], "description": "Models a lab has announced, teased or shown in a closed preview that nobody can download or call yet, with the day each was announced and when it's expected. We list a model when it comes out, so this is what we're watching. 13 waiting.", "facts": [ "13 waiting", "announced and expected dates", "sourced" ], "h1": "Coming models", "image": "https://www.anchorterminal.com/assets/og/upcoming.png", "path": "/upcoming/", "published": "", "section": "upcoming", "title": "Coming models: announced AI models not released yet | Anchor Terminal", "toc": null, "updated": "2026-10-23", "url": "https://www.anchorterminal.com/upcoming/" }, "tokens": { "markdown": 2100, "slim": 380 }, "version": 1 }