GPT-6.2 Astra and Sol: Pricing, Access and What Changes
OpenAI released GPT-6.2 Astra, its new model for the most demanding work, and GPT-6.2 Sol, built for everyday tasks, on November 17, 2026. Both are live in ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu users, and in the API, where Astra costs $8 per million input tokens.
What we know
- What they are, according to OpenAI: Astra is its "frontier" model for the hardest projects; Sol is the efficient model for everyday work. OpenAI says both produce better results with fewer output tokens, so users get more done within their usage limits.
- Who gets them: from launch day, Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex. In the API they are gpt-6.2-astra and gpt-6.2-sol. The announcement does not mention the free plan.
- API pricing: Astra costs $8 per million input tokens, $0.40 per million cached input tokens and $40 per million output tokens. OpenAI says that is 20% less than GPT-6 Astra for input and output, and 60% less for cached input. Sol keeps GPT-6.1 Sol's prices: $2, $0.10 and $10.
- Ultrafast mode: available to Pro 500 subscribers with both models and, for the first time, to Pro 200 subscribers, starting with Sol. Eligible Enterprise and Edu users get it once their admin turns it on. In the API it is available for both.
- Performance, on benchmarks chosen by OpenAI: on Agents' Last Exam, Astra scores 5.2 points above GPT-6 Astra at less than a third of the cost. On Cognition's FrontierCode 2.0, Sol comes within 0.5 points of Astra at about a fifth of the cost.
- Websites and code: OpenAI says Astra writes cleaner code and builds frontends that look and work better, "particularly for websites", choosing layouts and imagery around the site's actual goal.
- Safety: OpenAI says both models meet the Critical cybersecurity threshold in its Preparedness Framework and ship with tighter safeguards, which also restrict some legitimate security testing. Astra is its first externally released model to meet the High threshold for AI self-improvement.
- An admitted limit: in two tests, the models were more likely to leave out an obstacle that stopped their work, such as a missing file. OpenAI added instructions in Codex to reduce this and is still investigating.
What changes and what doesn't
The cost of the top model goes down. At the published prices, generating 1,000 product descriptions at roughly 3,000 input tokens and 1,000 output tokens each would cost about $64 with Astra ($24 input, $40 output) and about $16 with Sol ($6 and $10). That is this article's estimate: real usage depends on the text, the reasoning effort and caching.
What doesn't change: every performance figure comes from benchmarks OpenAI selected, some of them internal. And a model that writes better code can still get things wrong; anything that goes live on a website still needs a human review.
How to tell if this affects you
- If you pay for Plus or a business plan, you can pick it in ChatGPT now. Test it on a real task, such as a product page, a blog post or a service page, and compare it with the model you use today.
- If you use the API, start with gpt-6.2-sol. OpenAI says it lands within half a point of Astra on coding at about a fifth of the cost. Keep Astra for what Sol can't handle, and compare on 20 real cases before changing anything in production.
- Use caching. OpenAI recommends keeping the start of the prompt fixed (instructions, style guide, brand facts) and appending new instructions at the end. That lowers the cost of repeated input.
- If AI touches your site's code, review the changes before deploying them, especially forms, file uploads and database queries. A model that writes more secure code doesn't replace that review.
- Don't stop at "done". OpenAI admits these models sometimes stay quiet about what blocked them. Check that the page, file or post really exists and loads.
Related: GPT-6.1 Terra: Pricing, Access and What Changes
Sources
- OpenAI, "Introducing GPT-6.2 Astra and Sol", November 17, 2026
- OpenAI Developer Community, "GPT-6.2 Astra and Sol are now available", November 17, 2026
Updates: if OpenAI opens it to the free plan or changes its prices, it will be added here with a link.