ToolNavs Find Useful AI Tools
Submit Sign in
Back to AI information
GPT-6.2 Astra and Sol Are Here: Astra Costs 20% Less, Sol Nearly Matches the Flagship

GPT-6.2 Astra and Sol Are Here: Astra Costs 20% Less, Sol Nearly Matches the Flagship

AI information • Admin • • 33 views

On November 17, 2026, OpenAI released GPT-6.2 Astra and GPT-6.2 Sol. Both models are available starting today in ChatGPT Work and Codex, and through the API. It is the second major upgrade since the GPT-6 family debuted in September: Astra remains the flagship, Sol is the affordable option, and both use fewer output tokens for the same work — so you hit your usage limit later. Artificial Analysis already ranks GPT-6.2 second on its Intelligence Index, behind Claude Fable 5.2, which took the top spot on November 9.

Astra turns a screenshot into a template; Sol trails by half a point at a fifth of the price

Astra's changes center on two things. When you are building a slide deck, hand it a screenshot of a slide and it can complete the whole presentation to match the layout, colors, and even the font size. When editing an existing project in Codex, it writes cleaner code and more focused tests, and early testers report noticeably fewer rounds of revision. Sol's selling point is value: on Cognition's FrontierCode 2.0 benchmark, at maximum reasoning effort Sol trails Astra by less than half a point at about one-fifth of the cost. On OpenAI's own LifeVal test, Sol even beats the previous GPT-6 Astra at about one-seventh of the cost.

Pricing and availability were announced at the same time. Astra's standard input and output prices are 20% lower than GPT-6 Astra, while Sol keeps GPT-6.1 Sol pricing. Users on Plus, Pro, Enterprise, and Edu can use both models in ChatGPT Work and Codex starting today. The $200-a-month Pro plan gets Ultrafast mode on Sol for the first time; the $500-a-month Pro plan gets it on both models, and eligible Enterprise and Edu workspaces get it once an admin enables it. Ultrafast generates tokens up to 4.5 times faster than standard speed, but uses more of your allowance, and you switch speeds in the model picker.

What it means for developers

For developers, the practical effect is simple: allowances last longer and flagship capability costs less. People running long tasks in Codex will feel it first — the same monthly allowance covers considerably more work, and budget-conscious teams can let Sol handle most coding. There are two limits, though: the free tier is not in the first wave, and Ultrafast is reserved for the higher-priced plans, so regular Plus users cannot use it yet. On safety, an earlier hands-on test of GPT-6 Astra in network-attack scenarios raised its safe-response rate from 59% to 91.5%; whether this upgrade tightens the safety boundary further will depend on later reports.

What is worth watching next is real-world benchmarks and developer feedback: the official numbers look strong, but revision rates and stability on long tasks will only become clear once more people use the models. Both are callable through the API today; pricing and availability follow OpenAI's official documentation.

Recommended Tools

More