Reference
New AI models this week and this month
A dated list of model releases and updates from Anthropic, OpenAI, Google, xAI, DeepSeek and Mistral, read from the vendors' own model pages and changelogs. No leaderboard, no scores: a date, a model id, the vendor's one-line description and a link to the page that states it. Retirements come from the same vendors' deprecation pages. Subscribe by RSS to get the same list in your reader.
- Last 7 days
- 1
- Last 30 days
- 7
- Models tracked
- 65
- Checked
- Oct 28, 2026
Released in the last 30 days
| Date | Vendor | Model | Type | What the vendor says |
|---|---|---|---|---|
| Oct 26, 2026 | OpenAI | GPT-6.1 Terra gpt-6.1-terra | New model | API changelog: fast, efficient model for visual reasoning, computer use and writing (API date). Source ↗ |
| Oct 20, 2026 | Mistral | Voxtral 3 family | New model | Launched in Public Preview with four speech-to-text and text-to-speech models. Source ↗ |
| Oct 17, 2026 | DeepSeek | DeepSeek-V4.1-Flash-1017 deepseek-flash | New model | News page dated October 17, 2026: released on web, app, API and open-source (Hugging Face, ModelScope); the API id deepseek-flash is now served by it. Source ↗ |
| Oct 13, 2026 | Gemini 4 Argon gemini-4-argon | New model | Generally available release of the most capable model, built for complex coding, agentic, math and science tasks. Source ↗ | |
| Oct 7, 2026 | Anthropic | Claude Haiku 5.5 claude-haiku-5-5 | New model | Release notes: launched Claude Haiku 5.5, most capable model tuned for high-volume and latency-sensitive work. Source ↗ |
| Oct 6, 2026 | Gemini Nano Banana 2.1 gemini-nano-banana-2.1 | New model | Generally available image generation and editing model with better visual quality and aspect ratio support. Source ↗ | |
| Oct 6, 2026 | Mistral | Mistral Large 4 mistral-large-4 | New model | Entered Public Preview, with open weights promised later. Source ↗ |
Leaving the API next
| Retires | Vendor | Model | Replacement |
|---|---|---|---|
| Nov 30, 2026 | Anthropic | claude-sonnet-4-5-20250929 | claude-sonnet-5-5 |
| Dec 1, 2026 | OpenAI | gpt-image-1-mini, gpt-image-1.5, chatgpt-image-latest | gpt-image-2.5-sunburst or gpt-image-2.5-flare |
| Dec 11, 2026 | OpenAI | gpt-5, gpt-5-mini, gpt-5-nano, gpt-5-pro, o3, o3-pro | gpt-5.6-sol, gpt-5.6-terra or gpt-5.6-luna |
| Jan 6, 2027 | OpenAI | tts-1, tts-1-hd, gpt-4o-mini-tts | gpt-realtime-2.1-mini |
| Jan 14, 2027 | OpenAI | gpt-4o, gpt-4o-mini, gpt-4.1, gpt-4.1-mini (snapshots) | gpt-6-luna |
| Feb 26, 2027 | OpenAI | whisper-1, gpt-4o-transcribe, gpt-4o-mini-transcribe | gpt-live-transcribe or gpt-transcribe |
OpenAI: dated releases
Read from: developers.openai.com/api/docs/changelog · developers.openai.com/api/docs/models
| Date | Vendor | Model | Type | What the vendor says |
|---|---|---|---|---|
| Oct 26, 2026 | OpenAI | GPT-6.1 Terra gpt-6.1-terra | New model | API changelog: fast, efficient model for visual reasoning, computer use and writing (API date). Source ↗ |
| Sep 29, 2026 | OpenAI | GPT-6.1 Sol gpt-6.1-sol | New model | API changelog: lower-cost model for complex coding and professional work (API date). Source ↗ |
| Sep 22, 2026 | OpenAI | GPT-6 Sol gpt-6-sol | New model | API changelog: GPT-6 reasoning model for capability. Not listed on the models page at check time. Source ↗ |
| Sep 22, 2026 | OpenAI | GPT-6 Luna gpt-6-luna | New model | API changelog: GPT-6 reasoning model for lower cost (API date). Source ↗ |
| Sep 10, 2026 | OpenAI | GPT-Live 1 gpt-live-1 | New model | API changelog: full-duplex voice model, generally available in the API. Source ↗ |
| Sep 8, 2026 | OpenAI | GPT-Image-2.5 Sunburst gpt-image-2.5-sunburst | New model | API changelog: image generation and editing; gpt-image-2.5-flare released the same day. Source ↗ |
| Sep 3, 2026 | OpenAI | GPT-6 Astra gpt-6-astra | New model | API changelog: most capable model for end-to-end complex work (API date). Source ↗ |
| Aug 7, 2026 | OpenAI | GPT-5.6 Cyber gpt-5.6-cyber | New model | API changelog: security-focused model for approved defenders. Source ↗ |
| Jul 9, 2026 | OpenAI | GPT-5.6 Sol gpt-5.6-sol | New model | API changelog: GPT-5.6 family (sol, terra, luna); gpt-5.6 routes to Sol. Source ↗ |
| Jul 6, 2026 | OpenAI | GPT-Realtime-2.1 gpt-realtime-2.1 | New model | API changelog: updated realtime reasoning model; gpt-realtime-2.1-mini released the same day. Source ↗ |
| May 7, 2026 | OpenAI | GPT-Realtime-2 gpt-realtime-2 | New model | API changelog: realtime voice model, released with gpt-realtime-translate and gpt-realtime-whisper. Source ↗ |
| Apr 24, 2026 | OpenAI | GPT-5.5 gpt-5.5 | New model | API changelog: new frontier model; gpt-5.5-pro released the same day. Source ↗ |
| Apr 21, 2026 | OpenAI | GPT-Image-2 gpt-image-2 | New model | API changelog: new image generation and editing model. Source ↗ |
| Mar 17, 2026 | OpenAI | GPT-5.4 Mini gpt-5.4-mini | New model | API changelog: smaller, faster model for high-volume workloads; gpt-5.4-nano released the same day. Source ↗ |
| Mar 5, 2026 | OpenAI | GPT-5.4 gpt-5.4 | New model | API changelog: frontier model for professional work; gpt-5.4-pro released the same day. Source ↗ |
| Mar 3, 2026 | OpenAI | GPT-5.3 Chat gpt-5.3-chat-latest | New model | API changelog: points to the GPT-5.3 Instant snapshot used in ChatGPT. Source ↗ |
| Feb 24, 2026 | OpenAI | GPT-5.3 Codex gpt-5.3-codex | New model | API changelog: Codex model added to the Responses API. Source ↗ |
| Jan 14, 2026 | OpenAI | GPT-5.2 Codex gpt-5.2-codex | New model | API changelog: GPT-5.2 variant tuned for agentic coding added to the Responses API. Source ↗ |
| Dec 11, 2025 | OpenAI | GPT-5.2 gpt-5.2 | New model | API changelog: newest GPT-5 flagship, with better reasoning, coding and vision. Source ↗ |
| Dec 4, 2025 | OpenAI | GPT-5.1 Codex Max gpt-5.1-codex-max | New model | API changelog: Codex model for long-horizon agentic coding added to the Responses API. Source ↗ |
| Nov 13, 2025 | OpenAI | GPT-5.1 gpt-5.1 | New model | API changelog: released as the new GPT-5 flagship (API date). Source ↗ |
Mistral: dated releases
Read from: docs.mistral.ai/getting-started/changelog/ · docs.mistral.ai/getting-started/models/models_overview/
| Date | Vendor | Model | Type | What the vendor says |
|---|---|---|---|---|
| Oct 20, 2026 | Mistral | Voxtral 3 family | New model | Launched in Public Preview with four speech-to-text and text-to-speech models. Source ↗ |
| Oct 6, 2026 | Mistral | Mistral Large 4 mistral-large-4 | New model | Entered Public Preview, with open weights promised later. Source ↗ |
| Jul 16, 2026 | Mistral | OCR 4.1 mistral-ocr-4-1 | New model | Released; the mistral-ocr-latest and mistral-ocr-4 aliases now point to it (General Availability followed on 2026-08-31). Source ↗ |
| Jun 23, 2026 | Mistral | OCR 4 mistral-ocr-4-0 | New model | mistral-ocr-latest now points to it. Source ↗ |
| Apr 28, 2026 | Mistral | Mistral Medium 3.5 mistral-medium-3-5 | New model | Released as open weights under a Modified MIT license. Source ↗ |
| Mar 23, 2026 | Mistral | Voxtral TTS voxtral-tts-2603 | New model | Text-to-speech model with voice cloning and streaming. Source ↗ |
| Mar 16, 2026 | Mistral | Mistral Small 4 mistral-small-2603 | New model | Released together with Leanstral. Source ↗ |
| Dec 18, 2025 | Mistral | OCR 3 mistral-ocr-2512 | New model | OCR 3 released. Source ↗ |
| Dec 16, 2025 | Mistral | Mistral Small Creative labs-mistral-small-creative | New model | Released as a Labs model. Source ↗ |
| Dec 9, 2025 | Mistral | Devstral 2 devstral-2512 | New model | Released with Devstral Small 2. Source ↗ |
| Dec 2, 2025 | Mistral | Mistral Large 3 mistral-large-2512 | New model | Released together with the Ministral 3 models. Source ↗ |
DeepSeek: dated releases
Read from: api-docs.deepseek.com/news/
| Date | Vendor | Model | Type | What the vendor says |
|---|---|---|---|---|
| Oct 17, 2026 | DeepSeek | DeepSeek-V4.1-Flash-1017 deepseek-flash | New model | News page dated October 17, 2026: released on web, app, API and open-source (Hugging Face, ModelScope); the API id deepseek-flash is now served by it. Source ↗ |
Google: dated releases
Read from: ai.google.dev/gemini-api/docs/changelog · ai.google.dev/gemini-api/docs/models
| Date | Vendor | Model | Type | What the vendor says |
|---|---|---|---|---|
| Oct 13, 2026 | Gemini 4 Argon gemini-4-argon | New model | Generally available release of the most capable model, built for complex coding, agentic, math and science tasks. Source ↗ | |
| Oct 6, 2026 | Gemini Nano Banana 2.1 gemini-nano-banana-2.1 | New model | Generally available image generation and editing model with better visual quality and aspect ratio support. Source ↗ | |
| Sep 15, 2026 | Gemini 3.8 Live gemini-3.8-live | New model | Generally available audio-to-audio model for low-latency voice (an extended-thinking variant was released the same day). Source ↗ | |
| Sep 2, 2026 | Gemini 3.8 Flash gemini-3.8-flash | New model | Generally available release of the most intelligent Flash model, for long-horizon engineering and agents. Source ↗ | |
| Aug 13, 2026 | Gemini 3.7 Flash gemini-3.7-flash | New model | Generally available, aimed at coding and agent workflows, with an introductory price through December 31, 2026. Source ↗ | |
| Jul 21, 2026 | Gemini 3.6 Flash gemini-3.6-flash | New model | Generally available, with better token efficiency. Source ↗ | |
| Jul 21, 2026 | Gemini 3.5 Flash-Lite gemini-3.5-flash-lite | New model | Generally available, targets high-volume automation. Source ↗ | |
| May 19, 2026 | Gemini 3.5 Flash gemini-3.5-flash | New model | Generally available release of the model behind gemini-flash-latest, aimed at agentic and coding tasks. Source ↗ | |
| May 7, 2026 | Gemini 3.1 Flash-Lite gemini-3.1-flash-lite | Update | Generally available release of Gemini 3.1 Flash-Lite, optimized for speed, scale and cost. Source ↗ | |
| Mar 3, 2026 | Gemini 3.1 Flash-Lite Preview gemini-3.1-flash-lite-preview | New model | First Flash-Lite model in the Gemini 3 series. Source ↗ | |
| Feb 19, 2026 | Gemini 3.1 Pro Preview gemini-3.1-pro-preview | New model | Latest Gemini 3 Pro iteration, with a customtools variant that better prioritizes custom tools. Source ↗ | |
| Dec 17, 2025 | Gemini 3 Flash Preview gemini-3-flash-preview | New model | Fast, lower-cost frontier-class model with better visual, spatial and agentic coding abilities. Source ↗ | |
| Nov 18, 2025 | Gemini 3 Pro Preview gemini-3-pro-preview | New model | First Gemini 3 series model, a reasoning and multimodal model with agentic and coding strengths. Source ↗ |
Anthropic: dated releases
Read from: platform.claude.com/docs/en/release-notes/overview · platform.claude.com/docs/en/about-claude/models/overview
| Date | Vendor | Model | Type | What the vendor says |
|---|---|---|---|---|
| Oct 7, 2026 | Anthropic | Claude Haiku 5.5 claude-haiku-5-5 | New model | Release notes: launched Claude Haiku 5.5, most capable model tuned for high-volume and latency-sensitive work. Source ↗ |
| Sep 28, 2026 | Anthropic | Claude Sonnet 5.5 claude-sonnet-5-5 | New model | Release notes: launched Claude Sonnet 5.5; $2/$10 per MTok on overview. Source ↗ |
| Sep 22, 2026 | Anthropic | Claude Opus 5.5 claude-opus-5-5 | New model | Release notes: launched Claude Opus 5.5 for long-running agentic coding and knowledge work. Source ↗ |
| Sep 1, 2026 | Anthropic | Claude Fable 5.1 claude-fable-5-1 | New model | Release notes: launched Claude Fable 5.1 alongside Claude Mythos 5.1 for Project Glasswing participants; overview: for demanding reasoning and long-horizon agentic work, $10/$50 per MTok. Source ↗ |
| Jul 24, 2026 | Anthropic | Claude Opus 5 claude-opus-5 | New model | Release notes: launched Claude Opus 5, a step-change improvement over Claude Opus 4.8. Source ↗ |
| Jun 30, 2026 | Anthropic | Claude Sonnet 5 claude-sonnet-5 | New model | Release notes: launched Claude Sonnet 5 at introductory $2/$10 per MTok (standard price from Aug 10). Source ↗ |
| Jun 9, 2026 | Anthropic | Claude Fable 5 claude-fable-5 | New model | Release notes: launched Claude Fable 5, most capable model open to all customers, alongside Claude Mythos 5 for Project Glasswing participants. Source ↗ |
| May 28, 2026 | Anthropic | Claude Opus 4.8 claude-opus-4-8 | New model | Release notes: launched Claude Opus 4.8. Source ↗ |
| Apr 16, 2026 | Anthropic | Claude Opus 4.7 claude-opus-4-7 | New model | Release notes: launched Opus 4.7 at $5/$25 per MTok. Source ↗ |
| Apr 7, 2026 | Anthropic | Claude Mythos Preview | New model | Release notes: announced as gated research preview for defensive cybersecurity (Project Glasswing); not generally available. Source ↗ |
| Feb 17, 2026 | Anthropic | Claude Sonnet 4.6 claude-sonnet-4-6 | New model | Release notes: launched Sonnet 4.6, balanced model for everyday tasks. Source ↗ |
| Feb 5, 2026 | Anthropic | Claude Opus 4.6 claude-opus-4-6 | New model | Release notes: launched Opus 4.6, most intelligent model for complex agentic tasks. Source ↗ |
| Nov 24, 2025 | Anthropic | Claude Opus 4.5 claude-opus-4-5 | New model | Release notes: launched Claude Opus 4.5. Source ↗ |
xAI: dated releases
Read from: docs.x.ai/docs/release-notes · docs.x.ai/docs/models
| Date | Vendor | Model | Type | What the vendor says |
|---|---|---|---|---|
| Sep 2026 | xAI | Grok 4.7 grok-4.7 | New model | Release notes (September): available as grok-4.7, 500k context; models page gives knowledge cutoff May 2026. Day not stated. Source ↗ |
| Aug 2026 | xAI | Grok 4.6 grok-4.6 | New model | Release notes (August): available on the xAI API, 500k context. Day not stated. Source ↗ |
| Jul 2026 | xAI | Grok 4.5 grok-4.5 | New model | Release notes (July): priced $2/$6 per 1M tokens. Day not stated. Source ↗ |
| May 2026 | xAI | Grok Build 0.1 grok-build-0.1 | New model | Release notes (May): coding model for agentic workflows, early access. Day not stated. Source ↗ |
| Mar 2026 | xAI | Grok 4.20 and Grok 4.20 Multi-agent grok-4.20-0309-reasoning | New model | Release notes (March): both live; models page lists reasoning, non-reasoning and multi-agent ids, 1M context. Day not stated. Source ↗ |
| Nov 2025 | xAI | Grok 4.1 Fast | New model | Release notes (November 2025): available in the xAI Enterprise API. Day not stated. Source ↗ |
How this list is made
Where do the dates come from?
Only from pages the vendor publishes: model overview tables, API changelogs and release notes. A model appears here when one of those pages states a date. Announcements on social media, press coverage and leaderboards are not sources, and a model without a vendor-stated date is left out rather than dated by guess.
Why are there no scores or rankings?
Vendor benchmark numbers are self-reported and public leaderboards are regularly contested, so this page records what is verifiable: who released what and when. For a hands-on comparison on an identical task with the run conditions disclosed, see the agent benchmarks page.
How do I subscribe?
Add the RSS address above to any feed reader. The feed carries the same items as this page: releases, updates and retirements, each with the vendor source link. It is regenerated from the data file on every check date.
Does a release date mean the model is in the chat app?
Not necessarily. Dates here are API or developer-platform dates unless the note says otherwise. Consumer apps roll models out on their own schedule and by plan, which the vendor pages rarely state.