OpenAI upgrades ChatGPT and Codex with new GPT-6.2 Astra and Sol models
OpenAI began the GPT-6 generation with Astra in September. A few weeks later, at DevDay, it upgraded Sol to GPT-6.1.
Today, GPT-6.2 Astra and GPT-6.2 Sol are arriving together in ChatGPT Work and Codex. OpenAI says both models can complete difficult tasks using fewer output tokens. This means you can get more work done before reaching your usage limit.
What’s new in GPT-6.2 Astra and Sol?
Astra is still the model OpenAI recommends for its most difficult projects.
For example, imagine you have a screenshot of a slide and want the rest of your presentation to look the same. OpenAI says GPT-6.2 Astra can use the screenshot as a guide and match the layout, colors, and even font sizes.
In Codex, OpenAI says Astra is also better at making changes to existing projects and testing those changes.
“Early testers reported cleaner code style and more focused tests, with fewer rounds of revision,” OpenAI says.
Sol costs less and is designed for everyday coding and writing. It also comes surprisingly close to Astra in one coding test.
Cognition's FrontierCode 2.0 test checks whether changes to real software projects are good enough for the people who maintain them to accept and add them to the project.
OpenAI says that when both models used their highest reasoning setting, Sol scored within half a point of Astra while costing about one-fifth as much.
Some of you have been asking very politely. Some of you have just been asking.
— OpenAI (@OpenAI) November 17, 2026
GPT-6.2 Astra and Sol are now available.
They are very efficient and designed to work well on real-world tasks.https://t.co/qr2iP0GpfX pic.twitter.com/jLCyVIws2Y
Who gets GPT-6.2?
Starting today, both models are available in ChatGPT Work and Codex for Plus, Pro, Enterprise, and Edu subscribers.
The $200-per-month Pro plan also gets Ultrafast mode for the first time, but only with Sol. The $500-per-month Pro plan gets Ultrafast mode with both models. Eligible Enterprise and Edu workspaces can also use it after an administrator turns it on.
How fast is Ultrafast? OpenAI says it can produce tokens up to 4.5 times faster than the standard speed with either model. However, it also uses more of your usage allowance. You can change the speed in the model picker.
Developers can also use both models through the API starting today. Astra's standard prices for input and output are 20% lower than the prices for GPT-6 Astra. Sol keeps the same prices it had with GPT-6.1. You can find more details in OpenAI's full announcement.
