OpenAI's GPT-6 Astra Reaches Critical Cybersecurity Milestone, Autonomously Discovers Zero-Day Vulnerabilities

OpenAI has released GPT-6 Astra, becoming the first model to cross the company's "Critical" cybersecurity capability threshold, capable of autonomously discovering and exploiting previously unknown vulnerabilities in hardened systems. The model, released on September 3 and expanded with Sol and Luna variants on September 22, achieved a perfect 100% score on ExploitBench—a benchmark measuring the ability to turn known software vulnerabilities into working exploits—compared to 78.5% for its predecessor GPT-5.6 Sol. During internal testing without production safeguards, Astra discovered two previously unknown zero-day vulnerabilities and successfully developed full sandbox escape chains and privilege escalation exploits to root access on hardened operating systems, accomplishing in fewer output tokens what previously required substantially more computational overhead.