OpenAI Releases 'Astra' Cybersecurity Model Reaching 'Critical' Risk Threshold

charliermarsh · x · 2026-09-02

OpenAI has released 'Astra', a model designed for cybersecurity. Under the Preparedness Framework, Astra reaches the 'Critical' threshold. The company detailed the evaluation methodology, advanced safeguards, and future improvements. Astra is significantly more capable and token-efficient than 5.6 Sol, with advanced capabilities initially limited before broader access.

Related event: OpenAI Unveils Cybersecurity Model Astra Hitting Top Risk Threshold(2 posts)→

Original post →

More from Models

Models channel →