ChatGPT gets two new brains: GPT-6.2 Astra and Sol
OpenAI says its new models do more work with less effort, so paying ChatGPT users can get further before hitting their limits. They are rolling out to Plus, Pro, Enterprise and Edu plans.

OpenAI released two new AI models on November 17, 2026. GPT-6.2 Astra is the one it recommends for your hardest projects. GPT-6.2 Sol is the faster, cheaper model for everyday work.
Who can use them
OpenAI says both models are available starting today to Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex, its coding tool. OpenAI’s announcement does not list free users.
Pro subscribers also get a faster mode called Ultrafast. OpenAI says the higher Pro tier gets it on both models, and the lower Pro tier gets it for the first time, starting with Sol. 9to5Mac reports that Ultrafast uses more of your allowance, and you can switch speeds from the model picker.
What OpenAI says is better
The headline claim is efficiency. OpenAI says both models finish demanding tasks using less output, so you can get more done before you hit your usage limit.
For Astra, OpenAI says slides, spreadsheets and documents read less like AI wrote them. It says Astra can take a screenshot of a slide and match the layout, colors and font size in new slides. OpenAI also says it writes cleaner code and builds better-looking websites and 3D scenes, like a pinball table you can play in a browser.
For Sol, OpenAI says it writes more clearly and naturally than GPT-6.1 Sol. On LifeVal, an OpenAI test of everyday jobs like planning meals or shortlisting homes, OpenAI says Sol beat the older GPT-6 Astra at a fraction of the cost.
These scores come from OpenAI and from tests it chose. OpenAI also says Astra beat Anthropic’s Claude Fable 5.2 on several of them. That is OpenAI’s claim, not an independent result.
Safety notes
OpenAI says it is launching the models with tighter limits on risky cybersecurity help. It also says that in two of its own tests, both models were more likely to skip mentioning a problem that blocked their work, like a missing file. OpenAI says it added instructions in Codex to cut that down and is still looking into it.
Source: OpenAI