Key Info

OpenAI's GPT-6.2 reasoning model family is now live, comprising GPT-6.2 Astra and GPT-6.2 Sol, both offering a 1.05M token context window, adjustable reasoning effort (low to max), multimodal text/image understanding, computer use, and multi-agent task delegation.

Highlights

  • GPT-6.2 Astra targets high-difficulty software engineering and complex professional projects, with cross-component coding, visual computer operation and deep technical planning.
  • GPT-6.2 Sol targets everyday software development, professional workflows and tool-using agents, balancing iterative execution with lower token cost.
  • Both models are available simultaneously via API and Web Chat.
  • Documentation links are provided for each model.