SIAgentNews.com
← All news
DevelopingOpenAI

OpenAI puts GPT-6.2 Astra and Sol into ChatGPT Work and Codex

OpenAI's two newest models are live in ChatGPT Work, Codex and the API, promising finished tasks with fewer tokens and a lower price for the top model.

· First reported by 9to5mac.com

20%Astra's standard price cut versus GPT-6 Astra

The short version

  • GPT-6.2 Astra and Sol are available today in ChatGPT Work and Codex on Plus, Pro, Enterprise and Edu.
  • Astra is meant for the hardest projects; Sol is the cheaper pick for everyday coding and writing.
  • On Cognition's FrontierCode 2.0, Sol came within half a point of Astra at about one-fifth the cost.
  • Ultrafast mode generates tokens at up to 4.5 times the standard rate but uses more of your allowance.
  • Astra's standard API prices are 20% lower than GPT-6 Astra; Sol keeps its GPT-6.1 pricing.

Why it matters

If you already use ChatGPT Work or Codex, the new models arrive in the tools you have, and OpenAI says they stretch your usage limit further. The Astra-versus-Sol choice is also a practical money decision: Sol may be good enough for most daily work at a fraction of the cost.

What OpenAI released

OpenAI has launched GPT-6.2 Astra and GPT-6.2 Sol, two new versions of the models that power its products. As 9to5mac.com reported, they are rolling out to ChatGPT Work and to Codex, OpenAI's coding tool. Developers can also reach both through the API, the connection other software uses to talk to a model.

The release continues the GPT-6 generation. OpenAI started it with Astra in September, then moved Sol up to GPT-6.1 a few weeks later at its DevDay event. This update brings both models to the 6.2 level at the same time.

The headline claim is efficiency. OpenAI says both models can complete demanding tasks while producing fewer output tokens. Tokens are the small chunks of text a model reads and writes, and they are what your usage limit and the API bill are counted in. Fewer tokens per task means more finished work before you hit the cap.

Astra for hard jobs, Sol for daily ones

OpenAI still positions Astra as the model to pick for your toughest projects. Its example is a design one: give Astra a screenshot of a slide, and it can use that image as a template for the rest of a presentation, matching the layout, colors and even the font size.

In Codex, OpenAI says Astra is better at making and testing changes inside a project that already exists, which is much of real software work. The company says early testers saw cleaner code style and more focused tests, with fewer rounds of revision. That is OpenAI's account of tester feedback, not an independent check.

Sol is the more affordable option, aimed at everyday coding and writing. The most striking figure in the announcement involves Sol. On Cognition's FrontierCode 2.0, a test of whether changes to real codebases are good enough for the project's maintainers to accept, Sol came within half a point of Astra. That was with both models set to maximum reasoning effort, the setting that tells a model to think harder before answering. OpenAI says Sol did it at about one-fifth the cost.

Who gets it, and what Ultrafast means

Both models are available starting today in ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu plans. Which extras you get depends on your plan:

  • The $200/month Pro plan gets Ultrafast mode for the first time, but only with Sol.
  • The $500/month Pro plan gets Ultrafast on both Astra and Sol.
  • Eligible Enterprise and Edu workspaces get Ultrafast on both once an administrator turns it on.

Ultrafast is a speed setting. OpenAI says it can produce tokens at up to 4.5 times the standard rate on either model. The trade-off is that it draws down more of your allowance, so speed is something you spend. You switch between speeds from the model picker, the menu where you choose which model to use.

What changes for developers and budgets

For people building on the API, the pricing is the main news. Astra's standard input and output prices are 20% lower than they were for GPT-6 Astra. Sol keeps the pricing it had with GPT-6.1, so its cost advantage comes from being a cheaper model to begin with, not from a new discount.

Put together, the numbers point to a simple habit: try Sol first for routine work and move up to Astra when a task is hard or the stakes are high. That is our reading of OpenAI's figures, not a rule the company has stated. Other coverage today also notes that usage limits in Codex are worth checking before you move a large project over.

It is also worth keeping the claims in perspective. The benchmark result, the tester feedback and the token savings all come from OpenAI's own announcement. Your own results on your own work will tell you more than any single test.

What to watch

  • Do independent tests confirm OpenAI's claims about fewer tokens and Sol's close match to Astra?
  • How far does the extra usage from Ultrafast cut into plan allowances in everyday use?
  • Will OpenAI give clearer guidance on when to choose Astra over Sol?

Try it

Take a prompt you reuse often, such as a draft, a summary or a code change. Run it on Sol and then on Astra from the model picker, and compare quality against last month's result.

Sources and links

First reported by 9to5mac.com. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.

Liked it? Say why

Add a line for other readers.

No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.

More AI news