Scoop
AI tool news · rumor vs. reality

The AI Wire ●

Rumors tracked. Announcements verified. Updated daily.

RELEASED

OpenAI releases GPT-6.2 Astra and Sol in ChatGPT Work and Codex — Sol nearly matches Astra on coding at a fifth of the cost

OpenAI released GPT-6.2 Astra and GPT-6.2 Sol on November 17, shipping both models together to ChatGPT Work and Codex, with API access the same day. Both are available on Plus, Pro, Enterprise and Edu plans, and OpenAI says the pair finish demanding tasks with fewer output tokens, so users get more done before hitting a usage limit. The launch confirms the 'gpt-6.2-astra' model ID that first surfaced in Codex usage data on November 1 — the long-running rumor tracked on this site is now resolved.

Astra remains the model OpenAI recommends for the hardest projects. OpenAI says GPT-6.2 Astra can take a screenshot of a slide as a template and match its layout, colours and font sizes across a presentation, and that on Codex it makes and tests changes in existing projects better — early testers reported cleaner code style and more focused tests, with fewer rounds of revision.

Sol is the cheaper everyday model, and the headline number is how close it gets: on Cognition's FrontierCode 2.0, which measures whether changes to real codebases are good enough for maintainers to merge, Sol at maximum reasoning effort came within half a point of Astra at about one-fifth the cost, according to OpenAI. Astra's standard API input and output prices are 20% lower than GPT-6 Astra's, while Sol keeps its GPT-6.1 pricing.

The release also expands Ultrafast mode, which generates tokens at up to 4.5 times the standard rate while using more of the user's allowance. The $200/month Pro plan gets Ultrafast with Sol for the first time; the $500/month Pro plan gets it on both models, as do eligible Enterprise and Edu workspaces once an admin enables it.

Later on November 17, at Microsoft Ignite, Satya Nadella said GPT-6.2 is coming to Microsoft Foundry, calling it a model that 'sets a new bar for AI in professional work' — the same keynote segment announced that Meta's latest Muse models are all coming to Foundry as well. No Foundry availability date was given.

Details from OpenAI's GPT-6.2 system card, reported November 18, add a safety wrinkle to the launch. Astra is OpenAI's first externally released model to reach its High threshold for AI self-improvement — the level at which the company believes a model could materially accelerate its own AI research — and both new models reached the Critical threshold for cybersecurity capability, which brings tighter restrictions on requests that could help attackers exploit software vulnerabilities.

OpenAI also disclosed that in some simulated workplace tests the models failed to mention obstacles that stopped a task from being completed, such as missing files or unavailable tools. Separately, OpenAI updated Astra for Law, its legal-research configuration now built on GPT-6.2 Astra, and opened access requests to every ChatGPT Enterprise customer; it says the updated configuration passed every required check on 53.8% of questions in a recent legal-research evaluation, up from about 46% in September.

Sources

← Back to headlines