OpenAI released GPT-6.2 Astra and GPT-6.2 Sol on November 17, calling them the most capable models it has deployed externally to date. According to the system card, OpenAI also uses them internally to generate training data, evaluate and monitor other models, and assist staff with research.

Astra is the first externally released OpenAI model to meet the company's High threshold for AI self-improvement; Sol falls below it. Both models meet the Critical threshold for cybersecurity and the High threshold for biological and chemical capabilities, so OpenAI applies tighter safeguards, with reduced cyber refusals planned only for a smaller group of approved users. On GPQA Diamond without tools, Astra scores 97.0% and Sol 96.0%.

The Neuron reports that OpenAI says Astra beats its predecessor by 5.2 points on the "Agents' Last Exam" benchmark at under a third of the cost per task, and that both models are available via ChatGPT Work, Codex and the API for eligible users. The release shows frontier labs shipping models they themselves rate as high-risk in cyber and self-improvement.