OpenAI launches GPT-6.2 Astra and Sol, cutting prices and raising safety questions
By The Daily Byte
·20 Nov 2026
·5 min read

OpenAI has launched GPT-6.2 Astra and GPT-6.2 Sol, two new AI models that went live on 17 November 2026 in ChatGPT Work, the Codex coding tool and the company's API. Astra is the new flagship for the hardest projects, while Sol is the cheaper, everyday model for routine work.
The launch is the most significant model release from OpenAI in weeks, and it arrives with a price cut, a much larger context window and a safety rating that has drawn attention across the industry. Both models are also generally available on Amazon Bedrock from the same day, according to Amazon Web Services.
## What OpenAI released
OpenAI describes GPT-6.2 Astra as "our new frontier model for your hardest projects" and GPT-6.2 Sol as "our efficient, versatile model for everyday work". Astra is aimed at demanding work such as software development, scientific research and creating presentations, spreadsheets and documents. Sol is built for everyday tasks, writing and multi-step workflows.
Both models are available to Plus, Pro, Enterprise and Edu users inside ChatGPT Work and Codex. Developers can call them through the API as gpt-6.2-astra and gpt-6.2-sol. They are not in regular ChatGPT chat, and free users do not get access through ChatGPT Work.
Both models share a context window of 1,050,000 tokens, with up to 128,000 tokens of output. Astra's training data runs to 18 May 2026, while Sol's runs to 1 June 2026.
## Prices fall for Astra, hold steady for Sol
The headline price change is on Astra. Its standard API rates are $8 per million input tokens and $40 per million output tokens, with cached input at $0.40 per million. That is 20% below GPT-6 Astra, which cost $10 and $50, and cached input is 60% cheaper than before.
Sol keeps the rates of its predecessor, GPT-6.1 Sol: $2 per million input tokens, $0.10 for cached input and $10 per million output tokens. That makes Sol exactly a quarter of Astra's price.
OpenAI also added a faster Ultrafast mode for Pro subscribers. It is available on both models for Pro 500 subscribers, and on Sol for Pro 200 subscribers.
## A first for AI self-improvement
The safety picture is the part of the launch that has generated the most debate. Under OpenAI's Preparedness Framework, both models meet the Critical threshold for cybersecurity and the High threshold for biological and chemical capabilities.
Astra also becomes the first model OpenAI has released externally that reaches the High threshold for AI self-improvement — the company's measure of whether a model could help accelerate AI research itself. Sol does not reach that level.
OpenAI's own system card also records a mixed safety picture. The UK AI Security Institute reported that its testers built a universal cyber jailbreak against the new model using about half as many attempts as GPT-6 Astra had required. OpenAI says it believes it has fixed the problem, but the institute had not confirmed that at launch.
## What the benchmarks say — and what they do not
OpenAI's own figures put Astra ahead of its rivals on several tests. The company says Astra scores 5.5 points higher than Anthropic's Claude Fable 5.2 on Agents' Last Exam at about 5% of the cost, and leads on Artificial Analysis's AA-3D test with a score of 1,318, ahead of Fable 5.2 on 1,271 and Sol on 1,259.
Independent tracker Epoch AI placed Astra at 174 on its Capabilities Index, level with Claude Fable 5.2. Artificial Analysis ranks Astra second on its Intelligence Index, behind Fable 5.2, with Sol scoring 57 — matching the older GPT-6 Astra at roughly 15% of the cost.
These are vendor-reported and tracker figures, not independent laboratory tests, and early hands-on reactions have been mixed. Wharton professor Ethan Mollick, an early tester, described the model as uneven in his first impressions, while other testers praised its coding and document work.
## What it means for users
For businesses already paying for ChatGPT Work or Codex, the new models arrive automatically with no change in plan. The 20% price cut on Astra, combined with OpenAI's claim that it uses fewer output tokens per task than its predecessor, could lower the cost of heavy document and coding work.
For everyone else, the release is a reminder of how fast the frontier is moving: a flagship model launched in November is already cheaper than the one it replaced, and the safety questions around self-improvement are now attached to a model anyone with a paid plan can use. [Source: AI Magazine](https://aimagazine.com/news/gpt-6-2-astra-gpt-6-2-sol-openai-tops-the-pareto-frontier) and [TSN Media](https://tsnmedia.org/openai-gpt-6-2-astra-sol/), reporting OpenAI's launch announcement of 17 November 2026 and Amazon Web Services' Bedrock notice of the same day.
## Frequently asked questions
**When were GPT-6.2 Astra and Sol released?**
OpenAI released both models on 17 November 2026, in ChatGPT Work, Codex and the API.
**How much do GPT-6.2 Astra and Sol cost?**
Astra costs $8 per million input tokens and $40 per million output tokens. Sol costs $2 and $10. Both are standard short-context API rates.
**Can free ChatGPT users try GPT-6.2?**
No. The models are available in ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu users, not in regular ChatGPT or on free plans.