Model Release Tracker
AI model releases
Every notable AI model launch, newest first, with its context window and whether the weights are open.
| Date | Model | Context | Notes |
|---|---|---|---|
| 11 Nov | Gemini Decision 1 Google DeepMind | — | Decision model scoring up to 1,000 closed-ended questions in one pass across text, images, audio and video. |
| 11 Nov | AlphaGenome 2 Google DeepMind | — | Successor to AlphaGenome that predicts how DNA changes affect different cell types. |
| 11 Nov | Aure DeepL | — | Real-time voice translation model that carries over tone, emotion, emphasis and pace. |
| 10 Nov | Claude Mythos 5.2 Anthropic | — | Same model as Fable 5.2 with different safeguards, restricted to vetted cybersecurity and life sciences organizations. |
| 10 Nov | Gemini Robotics 2.5 Google DeepMind | — | Robotics model up to twice as successful at some tasks, released with an open safety benchmark. |
| 09 Nov | Claude Fable 5.2 Anthropic | 1M | Anthropic's most capable general-release model, up to 30% faster than Fable 5.1 and priced lower; Mythos 5.2 is the same model with fewer safeguards for vetted organizations. |
| 09 Nov | Muse Ember 1.0 Meta | 1M | Meta's smallest and most efficient Muse model, aimed at high-volume agent work like tool calls, routing and extraction. |
| 28 Oct | Gemini 4 Carbon Google DeepMind | — | Google's strongest model, focused on cybersecurity and long tasks, initially limited to vetted cyber defenders through the Fairwind Program. |
| 26 Oct | GPT-6.1 Terra OpenAI | — | Fast mid-tier GPT-6.1 model for visual reasoning and computer use at half the per-token price of GPT-6.1 Sol. |
| 23 Oct | Kimi K3.1 Pro OPENMoonshot AI | 1M | Post-trained upgrade of Kimi K3; its weights were published on Hugging Face in early November 2026 under a custom license. |
| 15 Oct | Muse Spark Open OPENMeta | 1M | Open-weight multimodal model (816B total, 42B active parameters) for coding and agentic work, under the Muse Model License. |
| 07 Oct | Claude Haiku 5.5 Anthropic | 1M | Anthropic's cheapest and fastest model, about 75% cheaper on average than Haiku 4.5. |
| 06 Oct | Mistral Large 4 OPENMistral AI | — | Natively multimodal 1-trillion-parameter model trained in Europe; launched as an API preview, with open weights following on October 29. |
| 30 Sept | Gemini 4 Argon Google DeepMind | — | First Gemini 4 model, aimed at software engineering, enterprise knowledge work and cyber defense, with a phased rollout. |
| 28 Sept | Claude Sonnet 5.5 Anthropic | 1M | Faster, lower-cost companion to Opus 5.5, at the same per-token price as Sonnet 5 but needing fewer tokens per task. |
| 22 Sept | Claude Opus 5.5 Anthropic | 1M | First Claude 5.5 model, performing near Fable 5.1 level at about 40% lower cost than Opus 5 on typical workloads. |
| 22 Sept | GPT-6 Sol and GPT-6 Luna OpenAI | — | Balanced and low-cost members of the GPT-6 family, priced at roughly half the GPT-5.6 models they replace. |
| 21 Sept | Grok 4.7 xAI | 500K | xAI's most capable model for coding and knowledge work, built on a larger base model than Grok 4.6. |
| 10 Sept | DeepSeek V4.1 Flash OPENDeepSeek | 1M | Natively multimodal MIT-licensed model that replaced V4 Pro on DeepSeek's API, with lower prices. |
| 03 Sept | GPT-6 Astra OpenAI | 1.05M | OpenAI's most capable model for complex reasoning, coding and computer use, rolled out in stages. |
| 03 Aug | Qwen3.8-Max OPENAlibaba (Qwen) | 1M | Alibaba's largest model (2.4T parameters), the first Max-class Qwen with downloadable weights, released as a text-only checkpoint. |
| 24 Jul | Claude Opus 5 Anthropic | 1M | First Claude Opus model of the 5 generation, since succeeded by Opus 5.5. |
| 09 Jun | Claude Fable 5 Anthropic | 1M | Anthropic's first Mythos-class model for general use, with safeguards that route some sensitive requests to another model. |
| 03 Jun | Gemma 4 12B OPENGoogle DeepMind | — | Apache 2.0 open model with native audio input that runs locally with 16GB of memory. |