AI News
Written by Joshua Reiher
By Joshua Reiher
Some links on this page are affiliate links. If you sign up through them we may earn a commission, at no extra cost to you.
![]()
Short answer: for new work, GPT-6.2 Astra is the better pick. OpenAI says it beats GPT-6 Astra on its professional-work, science and long-project tests, and in the API it costs 20% less per token. GPT-6 Astra is still worth knowing about because it remains in the ChatGPT model picker and still powers Pro reasoning in regular Chat, where GPT-6.2 Astra has not been added.
OpenAI released GPT-6.2 Astra on 17 November 2026 (US time) as the replacement for GPT-6 Astra, its previous top model. Here is how the two compare, using only OpenAI's own announcement, help centre, pricing and model pages, read on 18 November 2026. You can find more about the app itself on our ChatGPT listing.

Image: OpenAI
| GPT-6.2 Astra | GPT-6 Astra | |
|---|---|---|
| Released | 17 November 2026 | Before November 2026 (previous flagship) |
| Where in ChatGPT | Work and Codex only (not regular Chat), rolling out to Plus, Pro, Business, Enterprise, Edu | Work and Codex; also powers Pro reasoning in Chat |
| API price (per million tokens) | US$8 input, US$0.40 cached, US$40 output | US$10 input, US$1 cached, US$50 output |
| Context window | 1,050,000 tokens | 1,050,000 tokens |
| Max output | 128,000 tokens | 128,000 tokens |
| Knowledge cutoff | 18 May 2026 | 30 April 2026 |
| Best for | Documents, slides, spreadsheets, coding and computer use where quality matters | Existing workflows already tuned for it, and Pro reasoning in regular Chat |
OpenAI calls GPT-6.2 Astra its "new frontier model for your hardest projects". Its focus is professional output: OpenAI says its slides, spreadsheets and documents "read less like AI" and follow your templates and the conventions of fields like law, finance and accounting more closely. It can copy a slide template's layout, colours and font sizes from a file or a screenshot.
OpenAI also says it writes cleaner code with more focused tests, builds better websites and 3D experiences, and is more precise when using a computer. Its developer guide notes it is more concise by default and may ask before continuing with an unclear task.
In ChatGPT, it is rolling out in Work and Codex for Plus, Pro, Business, Enterprise, Edu, ChatGPT for Healthcare and FedRAMP workspaces. OpenAI says this rollout does not add it to regular Chat. On the Pro US$500 tier, it can run in the faster Ultrafast mode.
GPT-6 Astra was OpenAI's most capable model before this launch. OpenAI's model page describes it as built for complex reasoning, coding, computer use, research and document creation, with reasoning levels from low up to max. It is also available through Amazon Bedrock.
In ChatGPT, OpenAI's pricing page says Pro reasoning is powered by GPT-6 Astra. OpenAI's help centre says Pro US$100 and US$200 plans can use their full Work and Codex allowance on it, while Plus includes limited GPT-6 Astra usage, with optional credits after that.
All of these figures come from OpenAI's own announcement:
OpenAI also showed side-by-side examples, such as a valuation workbook and a slide deck, where it says GPT-6.2 Astra laid out the key figures more clearly than GPT-6 Astra. These are OpenAI's own tests, not independent results.
For developers, GPT-6.2 Astra is cheaper per token: US$8 input and US$40 output per million tokens, against US$10 and US$50 for GPT-6 Astra. Cached input drops from US$1 to US$0.40. Both charge more for requests over 272,000 input tokens. Because OpenAI says the new model also uses fewer output tokens on many tasks, the real saving can be larger than the per-token gap.
For ChatGPT users, there is no separate fee for either model. Both draw on your plan's included Work and Codex allowance. OpenAI lists Plus at US$20 a month and Pro from US$100 a month.
We used only OpenAI's official pages, read on 18 November 2026: the GPT-6.2 announcement (https://openai.com/index/introducing-gpt-6-2-astra-sol/), the ChatGPT Work and Codex help article (https://help.openai.com/en/articles/20001275-chatgpt-work-and-codex), ChatGPT pricing (https://chatgpt.com/pricing/), API pricing (https://developers.openai.com/api/docs/pricing), the model pages for GPT-6.2 Astra (https://developers.openai.com/api/docs/models/gpt-6.2-astra) and GPT-6 Astra (https://developers.openai.com/api/docs/models/gpt-6-astra), and the latest model guide (https://developers.openai.com/api/docs/guides/latest-model). We have not run our own tests.
No. OpenAI's rollout list covers Plus, Pro, Business, Enterprise, Edu, ChatGPT for Healthcare and FedRAMP workspaces, and only in Work and Codex.
Yes. OpenAI's help centre still lists GPT-6 Astra in Work and Codex, and its pricing page says it powers Pro reasoning. In existing Work chats, you can keep your current model when offered the switch.
Yes. It costs US$8 input and US$40 output per million tokens, compared with US$10 and US$50 for GPT-6 Astra, and cached input is 60% cheaper.
Already a member? The same buttons sign you in.
We'll email you a 6-digit code. No password needed. By continuing you agree to our Privacy Policy.
We sent a 6-digit code to . It expires in 10 minutes.
No email? Check your spam folder, or ·