OpenAI has launched GPT-6.2 Astra and GPT-6.2 Sol, its next generation of flagship models — and priced the bigger one to undercut its predecessor. The GPT-6.2 Astra price is $8 per million input tokens and $40 per million output tokens, with cached input at $0.40 per million: 20% less than GPT-6 Astra, and a 60% cut on cached input. Sol keeps GPT-6.1 Sol’s prices at $2 input and $10 output per million tokens. Announced on Monday, November 17, the pair arrives one week after Anthropic’s Claude Fable 5.2 took the top spot on the independent intelligence rankings — and OpenAI’s pitch is that Astra beats it on professional work at a fraction of the cost.
What GPT-6.2 Astra is
Astra is OpenAI’s new frontier model for the hardest projects: long professional documents, spreadsheets, slides, code and scientific research. OpenAI says its slides, spreadsheets and documents follow the templates and conventions of a field more closely than GPT-6 Astra’s did, and that it writes cleaner code and builds frontends that look and work better, particularly for websites, games and 3D design. Sol is the efficient everyday model — smarter and faster than GPT-6.1 Sol, built for multistep agent workflows, routine coding and writing, and the tasks people do outside work.
On OpenAI’s own benchmarks, Astra sets a new high on Agents’ Last Exam, a test of multistep professional assignments: 5.2 points ahead of GPT-6 Astra at less than a third of the cost per task, and 5.5 points ahead of Claude Fable 5.2 at just 5% of the cost. It also sets new highs on Terminal-Bench-Science 0.1 and the Vals Science Index, and on the legal test Harvey LAB-AA v1.1 it more than doubles Fable 5.2’s pass rate at about a sixth of the cost. On Artificial Analysis’s AA-3D test, where human judges compare interactive 3D experiences built from the same prompt, Astra ranks first at 1,318 Elo, ahead of Fable 5.2 at 1,271 and Sol at 1,259.
How to access it — and where it is not
Both models are available from today in ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu users, and in the OpenAI API as gpt-6.2-astra and gpt-6.2-sol. Ultrafast mode, which runs up to 4.5 times faster, is available with either model on the Pro $500 plan, and with Sol on Pro $200 — the first time Ultrafast has reached the $200 tier.
One thing to note: this rollout does not add either model to regular ChatGPT, which keeps GPT-6. To use Astra or Sol you need Work — ChatGPT’s agent mode for longer, multistep tasks — or Codex, or the API. Free and Go users do not get Work at all.
The independent verdict is mixed
OpenAI’s benchmark figures are its own, and the first independent score is cooler: Artificial Analysis scores Astra at 62 on its Intelligence Index, second behind Claude Fable 5.2, with Sol at 57. Astra climbs five points over GPT-6 Astra and overtakes Google’s Gemini 4 Carbon, but it does not retake the top spot. The cost story holds up better — Sol comes within 0.5 points of Astra on Cognition’s FrontierCode 2.0 at about a fifth of the cost.
The honest summary: OpenAI has closed the gap and cut the price of doing so, but on the independent index Fable 5.2 is still ahead. The models’ real advantage is value — near-frontier results at a small fraction of the cost per task.
Safety: a first, and a warning
Astra is OpenAI’s first externally released model to meet the High threshold for AI self-improvement under its Preparedness Framework — a measure of a model’s ability to accelerate AI research itself. Both models also meet the Critical threshold for cybersecurity and the High threshold for biological and chemical capabilities. OpenAI says it is deploying both with its strongest safeguards to date, including tighter restrictions on higher-risk cybersecurity assistance, and that defenders can access reduced-refusal versions through its Daybreak programme.
Why it matters
GPT-6.2 lands in the busiest month of flagship releases in years: Gemini 4 Hydrogen arrived on November 12, Claude Fable 5.2 on November 9, and Microsoft unveiled seven of its own MAI models on the same day as GPT-6.2. The pattern is now clear: the frontier labs are competing less on who has the smartest model and more on who delivers near-frontier work cheapest. OpenAI’s answer is a model that beats its rival’s flagship on the tests it chose, at 5% of the cost — and a cheaper sibling that does almost the same work for a fifth of the price.
What to watch
Three things will decide how far GPT-6.2 travels: independent benchmarks beyond Artificial Analysis, whether the models reach regular ChatGPT, and how Anthropic and Google answer on price.

One Comment