OpenAI has released GPT-6.2 Astra and GPT-6.2 Sol in ChatGPT Work and Codex on Plus, Pro, Enterprise and Edu, and in the API. OpenAI says both models finish demanding tasks with fewer output tokens, so you get more done before you hit your limits. Astra is still the one OpenAI recommends for the hardest jobs. It can match a slide deck’s layout, colors and fonts from a screenshot, and in Codex it’s better at making and testing changes in an existing project. OpenAI says early testers reported “cleaner code style and more focused tests.”
Sol is the cheaper everyday model. On Cognition’s FrontierCode 2.0, at max reasoning, OpenAI says it landed within half a point of Astra at about one-fifth the cost. Ultrafast mode, up to 4.5x the token rate in exchange for more of your allowance, comes to Sol on the $200 Pro plan and to both models on the $500 Pro plan and eligible Enterprise and Edu workspaces. In the API, Astra’s prices drop 20% from GPT-6 Astra, and Sol keeps GPT-6.1 pricing.
My take
A cheap model that comes within half a point on merge-worthy code is what I care about for daily agent work. I’ll try Sol in Codex on real tasks before I believe the benchmark, and save Astra for the gnarly stuff.
God Bless, and Use the Source!