OpenAI’s system card rates GPT-6.2 Astra at High for AI self-improvement, while GPT-6.2 Sol falls below that level, separating the models in the company’s safety assessment.
OpenAI’s card also places both models at Critical for cybersecurity and High for biological and chemical capabilities. OpenAI launched the pair Nov. 17; EconoTimes’ report on the launch covered the release.



