Live

Tech news, freshly booted.

AI

OpenAI releases GPT-6.2 Astra and Sol for ChatGPT, Codex and its API

The new versions of OpenAI's top model and its cheaper everyday model are rolling out to paying ChatGPT and Codex users today, with lower API pricing for Astra and a faster "Ultrafast" mode on more plans.

OpenAI has released GPT-6.2 Astra and GPT-6.2 Sol, updated versions of its flagship model and its lower-cost workhorse. In an announcement posted to its developer forum on Tuesday, the company said both models are available starting today in ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu subscribers, and to developers through the API.

The two models

OpenAI describes GPT-6.2 Astra as a frontier model for the hardest projects. The company says it sets new highs on its evaluations of professional work and scientific research, at less than a third of the cost per task of GPT-6 Astra, and that the improvement shows in work like slideshows, websites and 3D designs. 9to5Mac reports that Astra can take a screenshot of a slide and use it as a template, matching its layout, colors and font sizes, and that OpenAI says it is better at making and testing changes inside existing code projects in Codex.

GPT-6.2 Sol is the efficient everyday option. OpenAI says it handles multistep agent workflows, writes clearly and naturally, and is the go-to model for many of its own employees. According to 9to5Mac, OpenAI says Sol came within half a point of Astra on Cognition's FrontierCode 2.0, a test of whether changes to real codebases are good enough for maintainers to merge, at about one-fifth of the cost. Both models were run at maximum reasoning effort for that comparison.

OpenAI also says both models can finish demanding tasks with fewer output tokens, which 9to5Mac notes means subscribers can get more done before they hit usage limits.

Prices and plans

In the API, developers can call gpt-6.2-astra and gpt-6.2-sol. 9to5Mac reports that Astra's standard input and output prices are 20% lower than GPT-6 Astra's, while Sol keeps the pricing of GPT-6.1 Sol. Neither source lists per-token prices.

The bigger change for subscribers is speed. Ultrafast mode, which OpenAI says can generate tokens at up to 4.5 times the standard rate while using more of a user's allowance, is spreading to more plans:

  • Pro 500 subscribers can use Ultrafast with either model.
  • Pro 200 subscribers get Ultrafast for the first time, starting with GPT-6.2 Sol.
  • Enterprise and Edu workspaces can turn it on once an admin enables it, according to 9to5Mac.
  • API developers have Ultrafast available for both models.

Users can switch speeds from the model picker, 9to5Mac says.

How we got here

OpenAI started the GPT-6 generation with Astra in September, and Sol was later updated to GPT-6.1 at the company's DevDay event, according to 9to5Mac. The ".2" release keeps the same two-tier lineup but pushes on cost: OpenAI's pitch is that Astra does more for less, and that Sol now comes close to Astra on everyday coding.

Why it matters

The benchmark claims here are OpenAI's own, and neither source includes independent test results yet, so they're worth treating with some caution until outside evaluations appear. The practical changes are clearer. API customers who use Astra get a 20% price cut without switching models, Codex users get a cheaper model that OpenAI says is almost as good for code changes, and $200-a-month Pro subscribers get faster responses that were previously reserved for the more expensive plan. For teams choosing between the two, OpenAI's own numbers suggest trying Sol first and keeping Astra for the hardest work.

Sources

This story is based on reporting from these outlets. Read them for the full picture:

More from The Bootline