# OpenAI Unveils GPT-6 Cyber, Touting Enhanced Vulnerability Detection and Adherence > Source: > Published: 2026-11-05 17:54:13+00:00 OpenAI's latest GPT-6 Cyber model has made a significant leap in vulnerability detection, scoring 48.5 percent in its ExploitGym training evaluation - a substantial jump from its predecessor's 32.1 percent - while also being trained to operate within authorized boundaries. This critical model can detect and tackle zero-day flaws, but with a newfound emphasis on restraint.