OpenAI has released GPT-6.2 Astra and GPT-6.2 Sol to ChatGPT Work and Codex users.
The models are available to Plus, Pro, Enterprise, and Edu users, giving them a choice between Astra for difficult projects and Sol for everyday work.
In its announcement, OpenAI says both models produce better results on demanding tasks with fewer output tokens, potentially letting users do more within their usage limits.
Ultrafast access and model performance
Pro 500 subscribers can use Ultrafast with either model. Pro 200 subscribers get access for the first time with GPT-6.2 Sol. Eligible Enterprise and Edu users get the same access once their workspace admin enables it.
Ultrafast generates tokens at up to 4.5 times the standard rate, but consumes more of a user’s allowance, 9to5Mac reports. OpenAI’s announcement says Ultrafast is available for both models in the API at launch.
Astra is better at editing and testing existing code. Early testers reported cleaner code style and more focused tests with fewer rounds of revision. Astra can use a screenshot as a slide template, matching its layout, colours, and font size.
On Agents’ Last Exam, which tests multistep professional work, Astra scored 5.2 points higher than GPT-6 Astra at less than a third of the cost.
In Cognition’s FrontierCode 2.0, Sol came within 0.5 points of Astra at about one-fifth the cost, with both models at max. An independent ranking by Artificial Analysis puts Astra second on its Intelligence Index, behind Anthropic’s Claude Fable 5.2.
API pricing
GPT-6.2 Astra’s input and output API prices are 20% lower than GPT-6 Astra’s. Sol keeps the same prices as GPT-6.1 Sol; all prices below are per million tokens.
| Model | Input | Cached input | Output |
|---|---|---|---|
| GPT-6.2 Astra | $8 | $0.40 | $40 |
| GPT-6.2 Sol | $2 | $0.10 | $10 |
Developers can access the models as gpt-6.2-astra and gpt-6.2-sol. OpenAI introduced Sol and Luna API pricing in September.
Cyber safeguards and a coding regression
Under OpenAI’s Preparedness Framework, both models are rated Critical for cybersecurity and High for biological and chemical capabilities. Astra is also rated High for AI Self-Improvement, the first externally released OpenAI model to receive that rating.
The models have tighter default restrictions on higher-risk cybersecurity assistance, which OpenAI acknowledges can limit legitimate vulnerability research and security testing. Reduced cyber refusals for Astra are reserved for separately approved Daybreak Red organisations, continuing the access controls around Astra’s cyber capabilities.
GPT-6 Astra launched on 3 September, and OpenAI scrapped GPT-6.1 Astra later that month after it fell short on staying within bounds. Sol was upgraded to GPT-6.1 at DevDay a few weeks after Astra’s launch.
OpenAI’s system card also identifies a regression: GPT-6.2 Astra omitted a known coding obstacle in 22.28% of Blocker Deception trials, compared with 11.09% for GPT-6 Astra. Codex instructions substantially reduce those omissions.


















