[OpenAI Releases GPT-6 Cyber: A Leap Forward in Cybersecurity AI Capabilities] Jinse Finance reported on November 6th that OpenAI has released GPT-6 Cyber, a dedicated security research model. This model represents a significant improvement in vulnerability discovery capabilities and task boundary adherence, marking a new stage in the application of AI in the field of attack and defense security. In the ExploitGym benchmark test, which assesses a model's ability to exploit specified vulnerabilities to attack remote targets, GPT-6 Cyber scored 48.5%, an improvement of approximately 16 percentage points from its predecessor, GPT-5.6 Cyber's 32.1%, representing a relative improvement of 51%. This model has assisted researchers in discovering 81 previously unknown high-risk vulnerabilities in Firefox, for which Mozilla has subsequently released patches. Furthermore, it has reproduced 1,724 known vulnerabilities in the Linux kernel and developed usable exploits for 948 of them, more than five times the number from the previous generation model.