AI model release history.

224 major releases from OpenAI, Anthropic, Google, DeepSeek, Alibaba, Z.ai and Moonshot AI — 134 of them new models — dated from the makers’ own announcements. The newest model recorded here is GPT-6 Sol and GPT-6 Luna (22 Sep 2026).

224 MAJOR RELEASES7 LABSCHECKED 23 SEP 2026

NEWEST FROM OPENAI: GPT-6 Sol and GPT-6 Luna · 22 Sep 2026
NEWEST FROM ANTHROPIC: Claude Opus 5.5 · 22 Sep 2026
NEWEST FROM GOOGLE: Gemini 3.8 Flash · 2 Sep 2026
NEWEST FROM DEEPSEEK: DeepSeek-V4.1-Flash · 10 Sep 2026
NEWEST FROM ALIBABA: Qwen3.8-Flash-Next · 26 Aug 2026
NEWEST FROM Z.AI: GLM-5.3-Flash · 26 Aug 2026
NEWEST FROM MOONSHOT AI: Kimi K3 · 16 Jul 2026

§ 01 — LATEST

The latest major LLM and AI model releases.

ACROSS ALL SEVEN LABSNEWEST FIRST

GPT-6 Sol and GPT-6 Luna22 SEP 2026 · OPENAIOpenAI added GPT-6 Sol and GPT-6 Luna, cheaper GPT-6 models trained with similar methods to Astra, in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise and Edu. Free and Go users can use Luna in the desktop app, and the API models are gpt-6-sol and gpt-6-luna. They are not yet in standard Chat, and API standard pricing is US$2/US$10 per million tokens for Sol and US$0.10/US$0.50 for Luna.Model Claude Opus 5.522 SEP 2026 · ANTHROPICClaude Opus 5.5 launched as the first model in the Claude 5.5 family, at $4/$20 per million input/output tokens (Opus 5 was $5/$25), with a 1M token context window and always-on adaptive thinking. Anthropic said Sonnet 5.5 and Haiku 5.5 would follow in the coming weeks.Model Cowork and chat merged into one Claude16 SEP 2026 · ANTHROPICAnthropic merged Claude Cowork and chat into a single Claude interface and launched Claude Docs and Claude Slides, with Claude Design also usable inside conversations. The merge is rolling out to Pro and Max first, with Team and Free to follow and Enterprise admins given at least 30 days' notice. Docs, Slides and Design are in beta on paid plans.Product Agents API (public beta)10 SEP 2026 · OPENAIOpenAI released the Agents API in public beta for all developers. It provides the same managed harness and infrastructure that runs Codex, with durable sessions, automatic context compaction and a choice of OpenAI-hosted or third-party sandboxes. There is no extra fee beyond token and tool usage.Platform DeepSeek-V4.1-Flash10 SEP 2026 · DEEPSEEKDeepSeek-V4.1-Flash launches on the API as the smallest model in a new architecture family: a 552B-parameter causal encoder–decoder with 8B active parameters for input and 16B for output, native vision, and lower API prices. It replaces V4-Flash and the experimental V4-Flash-Vision-Exp.Model Gemini app for Windows10 SEP 2026 · GOOGLEGoogle launched a native Gemini app for Windows 10 and 11, available globally. It opens with Alt + Space over any window and connects to Gmail and Drive. Some features, such as Gemini Spark and Omni video generation, need a Google AI subscription.Product ChatGPT Images 2.58 SEP 2026 · OPENAIOpenAI launched Images 2.5 in ChatGPT, with sharper detail, more precise multi-turn editing and up to 50% lower generation latency than Images 2.0, along with Sketch, templates and on-image comments. In the API it released two models, GPT-Image-2.5 Flare for fast everyday generation and GPT-Image-2.5 Sunburst for precision editing.Model GPT-6 Astra3 SEP 2026 · OPENAIOpenAI released GPT-6 Astra, its first GPT-6 generation model, rolling out first to a limited set of organisations and then over the following days to Plus, Pro, Business and Enterprise users, the API, Microsoft Azure and AWS Bedrock. Pro, Business and Enterprise also get GPT-6 Astra Pro (shown as GPT-6 Pro in Chat), and API pricing is US$10 input / US$50 output per million tokens.Model Gemini 3.8 Flash2 SEP 2026 · GOOGLEGoogle released Gemini 3.8 Flash as generally available, its most capable Flash model, aimed at long-horizon software engineering and agents. It is more accurate than 3.7 Flash but uses more tokens, at the same introductory price of US$0.75/US$3.75 per million tokens. Gemini 3.8 Flash Cyber went to vetted defenders through the new Fairwind Program.Model Claude Fable 5.1 and Mythos 5.11 SEP 2026 · ANTHROPICClaude Fable 5.1 succeeded Fable 5 at the same $10/$50 per million token price, with cache reads cut to $0.25 per million. Claude Mythos 5.1 is the same model with lighter safeguards and is available only through Anthropic's trusted access programs.Model

§ 02 — THE TIMELINES

Pick an AI lab.

RELEASE HISTORIES7 TIMELINES

ChatGPT & OpenAI model release history54 RELEASES SINCE 2022OpenAI’s models and ChatGPT, from the November 2022 research preview to GPT-6. Latest model: GPT-6 Sol and GPT-6 Luna (22 Sep 2026).Timeline →Anthropic Claude release history50 RELEASES SINCE 2023Anthropic’s Claude models and apps, from the March 2023 launch to Claude 5. Latest model: Claude Opus 5.5 (22 Sep 2026).Timeline →Google Gemini release history43 RELEASES SINCE 2023Google’s Gemini models and apps, from Bard in February 2023 to Gemini 3.8. Latest model: Gemini 3.8 Flash (2 Sep 2026).Timeline →DeepSeek release history17 RELEASES SINCE 2023DeepSeek’s open-weights models and apps, from DeepSeek LLM in November 2023 to V4.1-Flash. Latest model: DeepSeek-V4.1-Flash (10 Sep 2026).Timeline →Alibaba Qwen release history30 RELEASES SINCE 2023Alibaba’s Qwen models and apps, from the Tongyi Qianwen launch in April 2023 to Qwen3.8. Latest model: Qwen3.8-Flash-Next (26 Aug 2026).Timeline →Z.ai GLM release history19 RELEASES SINCE 2023Z.ai’s (formerly Zhipu AI) GLM models and apps, from ChatGLM in March 2023 to GLM-5.3. Latest model: GLM-5.3-Flash (26 Aug 2026).Timeline →Kimi release history11 RELEASES SINCE 2025Moonshot AI’s Kimi models and agents, from Kimi k1.5 in January 2025 to the open 2.8T-parameter Kimi K3. Latest model: Kimi K3 (16 Jul 2026).Timeline →

§ 03 — SIDE BY SIDE

Every AI model release, seven labs.

CURRENTSUPERSEDEDRETIREDLARGER SQUARE = WATERSHED

§ 04 — PACE

AI model releases per year.

Each announcement counts once, even when it covers several models (Claude 3’s Opus, Sonnet and Haiku are one release).

MODEL RELEASES BY YEAR

Major model releases per year by lab
YEAR	OPENAI	ANTHROPIC	GOOGLE	DEEPSEEK	ALIBABA	Z.AI	MOONSHOT AI
2026	10	10	7	4	8	6	4
2025	10	6	8	5	9	6	3
2024	3	4	4	5	5	1	0
2023	2	3	3	1	3	3	0
2022	1	0	0	0	0	0	0

§ 05 — QUESTIONS

LLM and AI model releases, answered.

What is the newest AI model recorded for each lab?

OpenAI: GPT-6 Sol and GPT-6 Luna (22 Sep 2026), checked 23 Sep 2026. Anthropic: Claude Opus 5.5 (22 Sep 2026), checked 23 Sep 2026. DeepSeek: DeepSeek-V4.1-Flash (10 Sep 2026), checked 23 Sep 2026. Google: Gemini 3.8 Flash (2 Sep 2026), checked 23 Sep 2026. Alibaba: Qwen3.8-Flash-Next (26 Aug 2026), checked 23 Sep 2026. Z.ai: GLM-5.3-Flash (26 Aug 2026), checked 23 Sep 2026. Moonshot AI: Kimi K3 (16 Jul 2026), checked 23 Sep 2026.

Which lab has released the most models in 2026?

OpenAI, with 10 major model releases so far in 2026, against 10 from Anthropic, 8 from Alibaba, 7 from Google, 6 from Z.ai, 4 from DeepSeek and 4 from Moonshot AI. Each announcement counts once, even when it covers several models.

What counts as a major release?

A new model, a new product or surface, a new plan tier, or the launch of a capability that changed how the product is used (memory, search, voice, agents). Incremental updates, settings, feature drops and regional rollouts are left out, and retirements are noted on the model they retired.

Where do the dates come from?

Each date is the day the maker announced the release on its own site or in its own documentation, and every entry links to that source. The timelines are checked against the official release notes regularly and whenever a new model reaches our benchmark tables.

Dates are the day the maker announced each release, taken from their own site and documentation.

Major releases only: new models, products, plans and capability launches. Retirements are noted on the model they retired.

Checked 23 Sep 2026 · How the current models compare →

MAKE SENSE OF THE NEXT RELEASE

Choose AI improvements around your business needs.

DISCUSS YOUR AI OPTIONS AI SUPPORT & MAINTENANCE