AI Labs
Release tracker, by lab.
Every tracked model release and database change, plus sourced research highlights — filterable by lab and by event type. The full model-only timeline lives in the release tracker tool.
Lab:
| Date | Lab | Model | Type | What changed |
|---|---|---|---|---|
| Oct 15, 2026 | Meta AI | Muse Wonder | Added to database | Muse Wonder (Meta) added to the database — released 2026-10-12 at $4/$20 per 1M tokens, LiveBench 78.1. |
| Oct 15, 2026 | Mistral | Mistral Large 4 | Benchmark update | Mistral Large 4 LiveBench score corrected to 47.7 per LiveBench 2.0 (livebench.ai), replacing the 71.8 carried from a stale mirror leaderboard; League Score recomputed. |
| Oct 14, 2026 | Microsoft AI | MAI-Thinking-1 | Added to database | MAI-Thinking-1 (Microsoft AI) added to the database — released 2026-06-02. |
| Oct 14, 2026 | Microsoft AI | MAI-Code-1-Flash | Added to database | MAI-Code-1-Flash (Microsoft AI) added to the database — released 2026-06-02. |
| Oct 14, 2026 | Cohere | Command A+ | Added to database | Command A+ (Cohere) added to the database — released 2026-05-20. |
| Oct 14, 2026 | Cohere | Command A | Added to database | Command A (Cohere) added to the database — released 2025-03-13. |
| Oct 14, 2026 | AI21 | Jamba Large 1.7 | Added to database | Jamba Large 1.7 (AI21 Labs) added to the database — released 2026-01-15. |
| Oct 14, 2026 | AI21 | Jamba Mini 2 | Added to database | Jamba Mini 2 (AI21 Labs) added to the database — released 2026-01-15. |
| Oct 14, 2026 | Amazon | Nova 2 Pro | Added to database | Nova 2 Pro (Amazon) added to the database — released 2025-12-02. |
| Oct 14, 2026 | Amazon | Nova 2 Lite | Added to database | Nova 2 Lite (Amazon) added to the database — released 2025-12-02. |
| Oct 14, 2026 | Thinking Machines | Inkling | Added to database | Inkling (Thinking Machines Lab) added to the database — released 2026-07-15. |
| Oct 14, 2026 | Thinking Machines | Inkling Small | Added to database | Inkling Small (Thinking Machines Lab) added to the database — released 2026-07-31. |
| Oct 13, 2026 | DeepMind | Gemini 4 Argon | Model release | Released — 1M context, $2.00/$10.00 per 1M, League Score 91. |
| Oct 13, 2026 | DeepMind | Gemini 4 Argon | Added to database | Gemini 4 Argon released by Google DeepMind at $2/$10 per 1M tokens — added to the database. |
| Oct 13, 2026 | DeepMind | — | Research | Gemini 4 Argon reaches global release — Announced 30 Sep 2026 and released globally 13 Oct, Gemini 4 Argon launched at introductory pricing of $2/$10 per 1M tokens, with 'Video-to-Sim-to-Real' robotics demos. (Launch reporting via secondary outlets; the model's database record is sourced from Google's official pricing.) |
| Oct 12, 2026 | Meta AI | Muse Wonder | Model release | Released — 1M context, $4.00/$20.00 per 1M, League Score 76. |
| Oct 7, 2026 | Anthropic | Claude Haiku 5.5 | Model release | Released — 1M context, $0.10/$0.50 per 1M, League Score 68. |
| Oct 7, 2026 | Anthropic | Claude Haiku 5.5 | Added to database | Claude Haiku 5.5 released by Anthropic — added to the database. |
| Oct 7, 2026 | OpenAI | — | Research | GPT-6 rollout adds 'Intelligent UI' — OpenAI's GPT-6 rollout brought an 'Intelligent UI' — interactive charts and tools rendered inside answers — alongside the GPT-6 family already tracked in the League database. |
| Oct 7, 2026 | Anthropic | — | Research | Claude Haiku 5.5 announced — Anthropic's cheapest, fastest small model: roughly 75% lower average running cost than Haiku 4.5, with a 1M-token context window. |
| Oct 6, 2026 | Mistral | Mistral Large 4 | Model release | Released — 1M context, $0.68/$2.09 per 1M, League Score 4. |
| Oct 6, 2026 | Mistral | Mistral Large 4 | Added to database | Mistral Large 4 released — added to the database. |
| Oct 6, 2026 | Mistral | — | Research | Mistral Large 4 ('Le Chonk') public preview — A 1.05T-parameter open-weight MoE with a 1M-token context, trained in Mistral's own EU datacenters; weights due end of October 2026. |
| Oct 6, 2026 | DeepSeek | — | Research | DeepSeek V4.1 Flash weights published under MIT — V4.1 Flash (released 10 Sep 2026) got its MIT-licensed weights published on ModelScope on 6 Oct 2026 — the full open-weight playbook, again. |
| Oct 5, 2026 | Amazon | — | Research | Nova 2.5 Sonic generally available on Bedrock — Amazon's speech-to-speech model went GA on Bedrock on 5 Oct 2026 with a 256K-token context window. |
| Oct 1, 2026 | Microsoft AI | — | Research | MAI-Transcribe-2-Streaming ships — The latest addition to the seven-model in-house MAI family launched at Build 2026 (June) — Microsoft's first full frontier push trained entirely from scratch. |
| Sep 29, 2026 | OpenAI | GPT-6.1 Sol | Model release | Released — 1.1M context, $2.00/$10.00 per 1M, League Score 94. |
| Sep 28, 2026 | Anthropic | Claude Sonnet 5.5 | Model release | Released — 1M context, $2.00/$10.00 per 1M, League Score 87. |
| Sep 27, 2026 | Perplexity | — | Research | Legacy Sonar API tiers retired for the Agent API — Perplexity retired its legacy Sonar API tiers on 27 Sep 2026 in favour of the Agent API, months after launching Perplexity Computer (Feb 2026), its agent-orchestration product. |
| Sep 22, 2026 | OpenAI | GPT-6 Luna | Model release | Released — 1.1M context, $0.10/$0.50 per 1M, League Score 56. |
| Sep 22, 2026 | Anthropic | Claude Opus 5.5 | Model release | Released — 1M context, $4.00/$20.00 per 1M, League Score 96. |
| Sep 21, 2026 | xAI | Grok 4.7 | Model release | Released — 500K context, $2.00/$6.00 per 1M, League Score 67. |
| Sep 21, 2026 | xAI | — | Research | Grok 4.7 released — xAI's 'most capable model for coding and knowledge work': $2/$6 per 1M tokens with a 500K-token context window. |
| Sep 11, 2026 | Moonshot | — | Research | Kimi K2.8 Preview released — The K2.8 Preview brought a 1M-token context window across all Kimi Code tiers, two months after Kimi K3 landed (16 Jul 2026). |
| Sep 10, 2026 | DeepSeek | DeepSeek V4.1 Flash | Model release | Released — 1M context, $0.30/$1.20 per 1M, League Score 75. |
| Sep 3, 2026 | OpenAI | GPT-6 Astra | Model release | Released — 1.1M context, $10.00/$50.00 per 1M, League Score 97. |
| Sep 2, 2026 | DeepMind | Gemini 3.8 Flash | Model release | Released — 1M context, $0.75/$3.75 per 1M, League Score 63. |
| Sep 2, 2026 | Meta AI | Muse Spark 1.3 | Model release | Released — 1M context, $1.25/$4.25 per 1M, League Score 69. |
| Sep 2, 2026 | Meta AI | — | Research | Muse Spark 1.3 released — Meta's proprietary frontier line (first Muse Spark shipped Apr 2026) hit 1.3 — ranked top-5 on LiveBench at the lowest cost per task among proprietary models. |
| Sep 1, 2026 | Anthropic | Claude Fable 5.1 | Model release | Released — 1M context, $10.00/$50.00 per 1M, League Score 97. |
| Aug 12, 2026 | Qwen | — | Research | Qwen3.8-Max open weights follow launch — The 2.4T-parameter MoE flagship launched 3 Aug 2026 — and, a first for a Max-class Qwen, its open weights followed around 12 Aug under a bespoke licence. |
| Aug 11, 2026 | NVIDIA | — | Research | Nemotron 3.5 Lightning released — A 30B open-weight agent model (3B active) shipped with weights, data and recipes under the OpenMDW-1.1 licence. |
| Aug 10, 2026 | Meta AI | Muse Glimmer | Model release | Released — 131.1K context. |
| Aug 9, 2026 | AI21 | — | Research | AI21 sunsets its first-party Jamba API — AI21 shut down the first-party Jamba API and AI21 Studio on 9 Aug 2026. The Jamba models live on via cloud marketplaces (AWS Bedrock, OpenRouter) — which is why this lab carries no first-party price in our database. |
| Aug 3, 2026 | Qwen | Qwen3.8-Max | Model release | Released — 1M context, $2.00/$6.00 per 1M, League Score 62. |
| Jul 31, 2026 | Thinking Machines | Inkling Small | Model release | Released — 262.1K context. |
| Jul 27, 2026 | SSI | — | Research | NVIDIA strategic partnership + investment announced — SSI's first move out of stealth in two years: a strategic partnership with NVIDIA (Bloomberg reported the investment at $5B). Still zero shipped products — by design. |
| Jul 24, 2026 | Anthropic | Claude Opus 5 | Model release | Released — 1M context, $5.00/$25.00 per 1M, League Score 92. |
| Jul 21, 2026 | DeepMind | Gemini 3.5 Flash-Lite | Model release | Released — 1M context, $0.30/$2.50 per 1M, League Score 70. |
| Jul 16, 2026 | Moonshot | Kimi K3 | Model release | Released — 1M context, $3.00/$15.00 per 1M, League Score 63. |
| Jul 15, 2026 | Thinking Machines | Inkling | Model release | Released — 262.1K context, League Score 43. |
| Jul 15, 2026 | Thinking Machines | — | Research | Inkling released — a 975B open-weight debut — Thinking Machines Lab's first model: a 975B-parameter open-weight (Apache 2.0) multimodal MoE. Inkling Small followed on 31 Jul 2026. |
| Jul 9, 2026 | OpenAI | GPT-5.6 Sol | Model release | Released — 1.1M context, $4.00/$20.00 per 1M, League Score 82. |
| Jul 9, 2026 | OpenAI | GPT-5.6 Terra | Model release | Released — 1.1M context, $2.00/$12.00 per 1M, League Score 74. |
| Jul 9, 2026 | OpenAI | GPT-5.6 Luna | Model release | Released — 1.1M context, $0.20/$1.20 per 1M, League Score 60. |
| Jun 30, 2026 | Anthropic | Claude Sonnet 5 | Model release | Released — 1M context, $2.00/$10.00 per 1M, League Score 74. |
| Jun 12, 2026 | Moonshot | Kimi K2.7 Code | Model release | Released — 262.1K context, $0.95/$4.00 per 1M. |
| Jun 4, 2026 | NVIDIA | Nemotron 3 Ultra 550B-A55B | Model release | Released — 1M context, League Score 57. |
| Jun 2, 2026 | Microsoft AI | MAI-Thinking-1 | Model release | Released — 1M context. |
| Jun 2, 2026 | Microsoft AI | MAI-Code-1-Flash | Model release | Released — 256K context, League Score 1. |
| May 29, 2026 | xAI | Grok Build 0.1 | Model release | Released — 256K context, $1.00/$2.00 per 1M. |
| May 20, 2026 | Cohere | Command A+ | Model release | Released — 128K context, League Score 0. |
| May 20, 2026 | Cohere | — | Research | Command A+ released — A 218B MoE released under full Apache 2.0 with native citation generation — the enterprise RAG play, fully open. |
| May 20, 2026 | Stability AI | — | Research | Stable Audio 3 released — An open-weight diffusion-transformer audio family — Stability's media-model line continues outside the LLM race this database tracks. |
| Apr 30, 2026 | xAI | Grok 4.3 | Model release | Released — 1M context, $1.25/$2.50 per 1M, League Score 74. |
| Apr 28, 2026 | Mistral | Mistral Medium 3.5 | Model release | Released — 256K context, $1.50/$7.50 per 1M, League Score 59. |
| Apr 24, 2026 | DeepSeek | DeepSeek V4 Pro (0813) | Model release | Released — 1M context, $1.32/$3.96 per 1M, League Score 63. |
| Apr 20, 2026 | Moonshot | Kimi K2.6 | Model release | Released — 262.1K context, $0.95/$4.00 per 1M, League Score 70. |
| Apr 2, 2026 | DeepMind | Gemma 4 31B | Model release | Released — 262.1K context, League Score 49. |
| Apr 2, 2026 | DeepMind | Gemma 4 26B A4B | Model release | Released — 262.1K context, League Score 40. |
| Mar 16, 2026 | Mistral | Mistral Small 4 | Model release | Released — 256K context, $0.15/$0.60 per 1M. |
| Mar 11, 2026 | NVIDIA | Nemotron 3 Super 120B-A12B | Model release | Released — 1M context. |
| Feb 19, 2026 | DeepMind | Gemini 3.1 Pro Preview | Model release | Released — 1M context, $2.00/$12.00 per 1M, League Score 52. |
| Feb 16, 2026 | Qwen | Qwen3.5-397B-A17B | Model release | Released — 262.1K context, League Score 61. |
| Jan 15, 2026 | AI21 | Jamba Large 1.7 | Model release | Released — 256K context, League Score 0. |
| Jan 15, 2026 | AI21 | Jamba Mini 2 | Model release | Released — 256K context. |
| Dec 15, 2025 | NVIDIA | Nemotron 3 Nano 30B-A3B | Model release | Released — 1M context, League Score 0. |
| Dec 2, 2025 | Mistral | Mistral Large 3 | Model release | Released — 256K context, $0.50/$1.50 per 1M. |
| Dec 2, 2025 | Amazon | Nova 2 Pro | Model release | Released — 1M context, $1.25/$10.00 per 1M. |
| Dec 2, 2025 | Amazon | Nova 2 Lite | Model release | Released — 1M context, $0.30/$2.50 per 1M. |
| Oct 15, 2025 | Anthropic | Claude Haiku 4.5 | Model release | Released — 200K context, $1.00/$5.00 per 1M, League Score 16. |
| Aug 7, 2025 | OpenAI | GPT-5 | Model release | Released — 400K context, $1.25/$10.00 per 1M, League Score 60. |
| Jul 23, 2025 | Qwen | Qwen3-Coder-480B-A35B-Instruct | Model release | Released — 262.1K context, League Score 41. |
| Apr 5, 2025 | Meta AI | Llama 4 Scout | Model release | Released — 10M context. |
| Apr 5, 2025 | Meta AI | Llama 4 Maverick | Model release | Released — 1M context, League Score 48. |
| Mar 13, 2025 | Cohere | Command A | Model release | Released — 256K context, $2.50/$10.00 per 1M. |