OpenAI launches GPT-6 Cyber, capable of identifying vulnerabilities, building exploits, and proposing fixes.
OpenAI has launched GPT-6 Cyber, a model focused on uncovering software security flaws. It can detect defects, construct usable exploits, and recommend patches. On the ExpitGym benchmark, it scored 48.5%, surpassing its predecessor GPT-5.6 Cyber at 32.1%. When no safety instructions are provided, it remains within bounds in 99.93% of cases.
Return to original text
A short introduction is retained here; the full text from the source website is not reposted. For specific wording and subsequent updates, please refer to the original article.
Open AIHOT original