OpenAI releases GPT-6.2 Astra and Sol, cutting Astra's price 20% to $8 per 1M input tokens
OpenAI released GPT-6.2 Astra and GPT-6.2 Sol on 17 November 2026. Astra costs $8 and $40 per 1M tokens, 20% below GPT-6 Astra; Sol stays at $2 and $10.
By Gürkan E.Published
Tech AI LatestOpenAI released GPT-6.2 Astra and GPT-6.2 Sol on 17 November 2026 in the API, ChatGPT Work and Codex. Astra, the top model, costs $8 per million input tokens and $40 per million output tokens, 20% below GPT-6 Astra and the same list price as Claude Fable 5.2. Sol keeps $2 and $10. The system card rates both Critical for cybersecurity and makes Astra the first deployed OpenAI model rated High for AI self-improvement.
Key facts
- Announced: 17 November 2026 (OpenAI announcement and system card)
- GPT-6.2 Astra: $8 per 1M input tokens, $0.40 cached input, $40 per 1M output tokens
- GPT-6.2 Sol: $2 per 1M input tokens, $0.10 cached input, $10 per 1M output tokens, unchanged from GPT-6.1 Sol
- Availability: API as gpt-6.2-astra and gpt-6.2-sol; ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu users
- Preparedness Framework: both Critical in cybersecurity and High in biology and chemistry; Astra High in AI self-improvement
- Independent ranking: Artificial Analysis puts Astra second on its Intelligence Index at 62, behind Claude Fable 5.2 at 65
What OpenAI released
OpenAI released two models on 17 November 2026, GPT-6.2 Astra and GPT-6.2 Sol, in its announcement. OpenAI describes Astra as its most capable model for complex professional work, scientific research and engineering, and Sol as the more affordable model for everyday tasks and coding. Both are available in the API as gpt-6.2-astra and gpt-6.2-sol, and to Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex. OpenAI's help center says the models are not added to regular ChatGPT chat, and that workspace access opens in stages.
Astra's API list price is $8 per million input tokens, $0.40 per million cached input tokens and $40 per million output tokens. That is 20% below the $10 and $50 that GPT-6 Astra launched at in September, and 60% below its $1 cached input rate. Sol keeps the $2 input, $0.10 cached input and $10 output prices of GPT-6.1 Sol. Coverage of the launch lists a 1,050,000-token context window for both models, the same as the GPT-6 generation.
What OpenAI claims on performance and cost
OpenAI's headline claim is efficiency rather than a large capability jump. The company says Astra scores 5.2 percentage points higher than the original GPT-6 Astra on Agents' Last Exam at a fraction of the cost per task. It also says Astra used 62% fewer tokens than GPT-6.1 Sol on complex legal agent tasks in its own evaluations. These are vendor figures on benchmarks OpenAI selected.
Independent results are mixed. Artificial Analysis ranks Astra second on its Intelligence Index with 62 points, behind Claude Fable 5.2 at 65, with Sol at 57, and it reports no improvement in coding agent performance over GPT-6 Astra. Vals AI places Astra seventh on its Vals Index. Android Authority reported that some early testers described unproductive loops and tasks marked complete too early, despite the benchmark gains.
What the system card says
The GPT-6.2 system card rates both models Critical for cybersecurity under OpenAI's Preparedness Framework, the level GPT-6 Astra first reached in September, and High for biological and chemical capability. Because of the Critical cyber rating, OpenAI says a version of its monitoring system watches external use of the models and that safeguards also apply to internal use. OpenAI also says it added tighter restrictions on higher-risk cybersecurity requests.
The new element is self-improvement. OpenAI rates Astra High for AI self-improvement, the first externally deployed model it has placed at that level; Sol is not rated High in that category. The card does not claim that Astra can improve itself without people. The High level is a threshold that, under OpenAI's own framework, requires added safeguards before deployment.
What it means for buyers
The price cut puts OpenAI's top model at exactly the list price of Anthropic's top general model. Claude Fable 5.2 launched on 9 November at $8 per million input tokens and $40 per million output tokens, also a 20% cut from its predecessor. Eight days later, the two flagship models cost the same per token, so the real difference for API customers is how many tokens each model spends on a task and how often it finishes the task correctly. That is why both companies now lead with cost per task instead of price per token.
For teams already on GPT-6.1 Sol, the move to GPT-6.2 Sol costs nothing extra per token. For teams on GPT-6 Astra, the same workload should cost about 20% less at list price before any token savings. The open questions are whether OpenAI's token savings hold on customers' own workloads, which the announcement cannot show, and whether the agent reliability complaints from early testers persist once the staged rollout reaches all workspaces. Our model release tracker lists both rows.
What changed
| Before | GPT-6 Astra: $10 input, $1 cached, $50 output per 1M tokens; GPT-6.1 Sol: $2 and $10; no deployed OpenAI model rated High for AI self-improvement |
|---|---|
| After | GPT-6.2 Astra: $8 input, $0.40 cached, $40 output per 1M tokens; GPT-6.2 Sol: $2 and $10; Astra rated High for AI self-improvement |
| Who it affects | API customers using Astra or Sol; Plus, Pro, Enterprise and Edu users of ChatGPT Work and Codex |
Questions people ask
- How much does GPT-6.2 Astra cost?
- $8 per million input tokens, $0.40 per million cached input tokens and $40 per million output tokens at OpenAI's list price, 20% below GPT-6 Astra.
- Is GPT-6.2 Sol cheaper than GPT-6.1 Sol?
- No. GPT-6.2 Sol keeps GPT-6.1 Sol's prices of $2 per million input tokens and $10 per million output tokens.
- Where can I use GPT-6.2 Astra and Sol?
- In the API as gpt-6.2-astra and gpt-6.2-sol, and in ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu users. OpenAI says they are not added to regular ChatGPT chat.
- What does the High self-improvement rating mean?
- It is a level in OpenAI's Preparedness Framework that requires extra safeguards before deployment. GPT-6.2 Astra is the first externally deployed OpenAI model rated High in that category.
Sources
- Introducing GPT-6.2 Astra and Sol (OpenAI, 17 Nov 2026)
- GPT-6.2 Astra and Sol system card (OpenAI, 17 Nov 2026)
- ChatGPT Work and Codex (OpenAI Help Center, 17 Nov 2026)
- OpenAI has the #2 model in the Intelligence Index with the release of GPT-6.2 Astra and GPT-6.2 Sol (Artificial Analysis, 17 Nov 2026)
- OpenAI upgrades ChatGPT and Codex with new GPT-6.2 Astra and Sol models (9to5Mac, 17 Nov 2026)
- OpenAI launches GPT-6.2 Astra and Sol with lower API prices and some benchmark gains (The Decoder, 17 Nov 2026)
- OpenAI launches GPT-6.2 Astra and Sol, but not everyone is sold on the cheaper upgrade (Android Authority, 18 Nov 2026)
Reference pages
- GPT-6.2 Astra model
- GPT-6.2 Sol model
- GPT-6 Astra model
- GPT-6.1 Sol model
- Claude Fable 5.2 model
- OpenAI company
- Codex tool
- ChatGPT tool
Spotted an error? See the corrections policy and tell the editor. Facts in this story come from the sources listed above; vendor claims are attributed.
Comments
Loading comments.





