HomeBriefFor youTrackersTopicsGlossaryNewsletterSecurityHealth & medicineFinance & marketsRoboticsModels & researchPolicy & regulationEnterprise & businessDeveloper tools
Tracker

AI model release tracker

Every notable model release from the major labs, newest first, with what it is for and where it is available.

Last updated 2026-11-08. Newest first; each entry links to the vendor page.

2026-11-05 · OpenAI
GPT-6 Cyber Specialised (security)
GPT-6 Astra tuned for authorised vulnerability discovery and exploit work. First OpenAI model rated Critical for cyber capability. Limited to vetted Daybreak Red customers. source
2026-11-04 · OpenAI
GPT-Image 2.6 (Sunburst / Flare) Image generation
Two variants: Flare for speed, Sunburst for quality. Available in the API and ChatGPT. (Reported; vendor changelog is the authority.) source
2026-10-29 · Mistral
Mistral Large 4 ("Le Chonk") Open-weight flagship
Open-weight release of Mistral's largest model. (Reported from Mistral's news page.) source
2026-10 (listed) · Google
Gemini 4 Argon Flagship
Google's current top model on the Gemini API. Launch pricing $2 / $10 per million tokens through Nov 30, then $4 / $20. source
2026 (listed) · Google
Gemini 3.8 Flash Fast / mid-tier
Current Flash tier with a free tier for testing; $0.75 / $3.75 through Dec 31, 2026. source
2026-09 · Anthropic
Claude Fable 5.1 Flagship (top tier)
Anthropic's most capable tier, above Opus. $10 / $50 per million tokens. Shares a base model with Mythos 5.1 but adds safety measures for biology, cyber and AI R&D. source
2026-09-04 · OpenAI
GPT-6 Astra Flagship
OpenAI's current general flagship. $10 / $50 per million tokens; 128K max output; knowledge cutoff April 2026. source
2026 (current lineup) · Anthropic
Claude Opus 5.5 / Sonnet 5.5 / Haiku 5.5 Flagship / mid / small
Current three-tier lineup below Fable. Opus 5.5 is priced below the previous Opus generation ($4 / $20 vs $5 / $25). source
2026-06-09 · Anthropic
Claude Fable 5 / Mythos 5 Flagship (top tier)
First Mythos-class models. Access was suspended June 12–30, 2026 under US export controls and restored July 1. source

Dates are the vendor's announced general or API availability date where published. Rows marked "listed" are models currently on the vendor's pricing page whose exact launch date was not stated there; the nightly check replaces these with announcement dates as they are confirmed.

Common questions

Which model is the most capable right now?

There is no single answer across tasks. The top tiers are OpenAI's GPT-6 Astra, Anthropic's Claude Fable 5.1 and Google's Gemini 4 Argon; vendor-published benchmarks favour the vendor publishing them, so compare independent evaluations for your specific workload.

What does 'open-weight' mean?

The model's trained parameters are published so anyone can download and run it, as Mistral did with Mistral Large 4. Closed models such as GPT-6 Astra are only reachable through the vendor's API.

How is this list kept current?

A nightly pass checks the major labs' announcement pages, API changelogs and pricing pages and adds any new release with a link to the vendor's own page.

Get the brief in your inbox

One email every morning: the stories that mattered in AI, each with a one-line "why it matters" for your role. Free, no ads, unsubscribe any time.

Published daily by AI Briefing. We never share your address.
Like this? Get tomorrow's brief by email, free.