OpenAI released two new models on Tuesday, Nov. 17. GPT-6.2 Astra is its new top model for long, complex work, and GPT-6.2 Sol is the cheaper model the company now calls its "go-to model for everyday work." Both are available starting today to ChatGPT Plus, Pro, Enterprise and Edu subscribers, but only in ChatGPT Work and Codex.
If you open the app, look in Work. The ChatGPT release notes say you pick Astra or Sol in the Work model picker, and that "this rollout does not add either model to regular Chat." The announcement doesn't list free accounts among the plans that get them.
Who gets what
| Plan | Access on Nov. 17 |
|---|---|
| Plus, Pro | Astra and Sol in ChatGPT Work and Codex |
| Pro 500 | Both models, plus the faster Ultrafast mode for either one |
| Pro 200 | Ultrafast for the first time, starting with Sol |
| Enterprise, Edu | Same access once a workspace admin turns it on |
| Developers (API) | gpt-6.2-astra and gpt-6.2-sol, with Ultrafast at launch |
API prices for Astra are $8 per million input tokens, $0.40 for cached input and $40 for output. OpenAI says that is 20% lower than GPT-6 Astra on regular tokens, and 60% lower on cached input. Sol keeps GPT-6.1 Sol's rates of $2, $0.10 and $10.
What OpenAI says changed

OpenAI's pitch is better office work for less money. It says Astra's slides, spreadsheets and documents follow a user's templates and the conventions of fields like law and finance more closely, and that its writing has "fewer generic takeaways and less filler." According to the release notes, Astra also does a better job matching the voice of a draft you ask it to edit.
OpenAI also published benchmark results. On Agents' Last Exam, a test of multistep professional tasks, OpenAI says Astra beat GPT-6 Astra by 5.2 points for under a third of the cost per task. It also claims Sol beats GPT-6 Astra on its own LifeVal test of everyday chores, like choosing kindergartens after a family moves, for roughly a seventh of the cost. All of those figures come from OpenAI.
Edit USA wrote about GPT-6 Astra's role in a particle physics proof earlier this month.
The safety ratings

OpenAI rates both models at its "Critical" level for cybersecurity under its Preparedness Framework, as it did GPT-6 Astra. They also reach its "High" level in biology and chemistry, and OpenAI says it added restrictions on some dual-use requests in those areas.
GPT-6.2 Astra is also OpenAI's first publicly released model to reach "High" for AI self-improvement, the company's measure of how much a model can speed up AI research. OpenAI describes that level as equal to giving every one of its researchers "an additional, highly capable research engineer as an assistant," compared with 2024. It says it added containment in its research environments and a system called Control Room that can halt a model's work when someone needs to step in.
OpenAI also flagged one behavior it hasn't fully fixed. In two of its tests, the new models were more likely than GPT-6 Astra to leave out an obstacle that blocked their work, like a missing file or a search tool that wasn't available. The company says new instructions in Codex substantially reduce those omissions and that it is still investigating.














