News
November 18, 2026 · OpenAI

OpenAI releases GPT-6.2 Astra and Sol, cutting Astra's API price

OpenAI launched two GPT-6.2 models on November 17: Astra for demanding professional work and Sol for fast everyday use. Astra's API price drops by a fifth.

OpenAI introduced two new models, GPT-6.2 Astra and GPT-6.2 Sol, on November 17, 2026. According to OpenAI's announcement and a detailed report by The Decoder, Astra targets specialised work in fields such as science, law and finance, while Sol is meant to be a quick, general model for daily tasks.

What's new

  • Two tiers: OpenAI positions Astra as its strongest model for professional output and Sol as the cheaper, faster option. The Decoder reports that on the FrontierCode 2.0 coding test, OpenAI says Sol comes within half a point of Astra at about a fifth of the cost.
  • Better office output: OpenAI says Astra follows professional conventions more reliably, for example by taking layouts, colours and fonts from a template when it builds slides.
  • Independent rankings: The Decoder cites Artificial Analysis, which places Astra second on its Intelligence Index with 62 points, behind Claude Fable 5.2 at 65, and puts Sol at 57.
  • Weak spots: the same outlet notes that Artificial Analysis saw no coding gain for Astra over its predecessor, and that Vals AI found the models ranked low on some legal research tests.
  • Safety: OpenAI says both models reached the "Critical" level for cybersecurity in its Preparedness Framework, so it is adding stricter safeguards for risky security requests.

Availability and pricing

Subscribers on the Plus, Pro, Enterprise and Edu plans can pick either model now, both inside ChatGPT Work and in Codex; developers get them through the API. According to The Decoder, Astra costs $8 per million input tokens, $0.40 for cached input and $40 per million output tokens, which OpenAI describes as 20 percent cheaper than before. Sol keeps its predecessor's price of $2 input, $0.10 cached and $10 output per million tokens.

Why it matters

Analysis: the headline here is price as much as capability. For teams paying per token, a cheaper top model and a Sol tier that sits close to it on coding make it worth re-running their own evaluations. The mixed outside results on coding and legal work are a reminder to test on real tasks before switching.

Source: OpenAI