What are GPT-6.2 Astra and GPT-6.2 Sol?
They are two new models from OpenAI, announced on November 17, 2026. OpenAI says they are its most capable externally deployed models to date.
"Externally deployed" refers to models released to users and developers outside the company. The claim covers both models, so Astra and Sol are presented as a pair rather than as a flagship and a footnote.
What is new in GPT-6.2?
OpenAI's research announcement describes improved professional outputs, coding, computer use, and efficiency at lower cost. It says those gains apply across ChatGPT Work, Codex and the API.
Four areas stand out:
- Professional outputs: the kind of documents and deliverables people produce for work.
- Coding: writing and changing software, including in Codex.
- Computer use: models operating software on a user's behalf.
- Efficiency at lower cost: more done for less spend.
The announcement does not tie these gains to one model over the other, so treat them as the release's overall direction.
Who can use GPT-6.2 Astra and Sol, and when?
Access started on announcement day. Both models were available in ChatGPT Work and Codex for Plus, Pro, Enterprise, and Edu users.
Developers did not have to wait either. Both models were accessible in the OpenAI API on announcement day.
If you use ChatGPT Work or Codex on one of those plans, you can try the models in the tools you already have. If you build your own products, you can call them through the API.
How much do GPT-6.2 Astra and Sol cost?
API pricing is per million tokens, and the two models sit at different price points. OpenAI's pricing page lists GPT-6.2 Astra at $8 per million input tokens and $40 per million output tokens.
On the same page, GPT-6.2 Sol is listed at $2 per million input tokens and $10 per million output tokens.
Output tokens cost more than input tokens on both models. Tasks that generate long responses, such as code generation or lengthy agent runs, will lean on the output rate.
How do Astra and Sol compare?
Price is the clearest difference. Astra is more expensive on both input and output, as the pricing page shows.
Positioning is the other. OpenAI's API changelog positions Sol for everyday coding and agent workflows. That makes Sol the natural starting point if your work is routine coding help or agents that make many calls, where token costs add up.
The materials describe Astra's role less specifically. If you are weighing it, test it against Sol on your own prompts and compare output quality against the higher bill.
Frequently Asked Questions
Which API endpoints work with these models?
Both models support the Responses API and Chat Completions. If your app already uses either, you can point it at the new models without switching interfaces.
Where should I check current details?
OpenAI's API changelog tracks model support and positioning. Its pricing page holds the current rates. Check both before you budget a project.



