According to CoinMeta, OpenAI recently released the GPT-6 Cyber model to customers who have passed the review process. This model is specifically designed to identify and exploit security vulnerabilities. The network security risk rating of this model has been raised from the previous "high" level to a "critical" level, with the aim of helping security researchers to discover vulnerabilities in advance and test potential exploitation methods. OpenAI Researchers have used this model to identify 81 unknown Firefox vulnerabilities, all of which have been classified as "high severity" by Mozilla and repairs have already begun. In addition, this model has assisted in developing exploits for 948 Linux kernel vulnerabilities (including 588 unknown vulnerabilities), and has disclosed 304 critical vulnerabilities along with patch recommendations. GPT-6 Cyber is currently available to specific customers and will be extended to all approved customers in the coming weeks. Meanwhile, OpenAI will make the early version of the GPT-5.6 Cyber models available to relevant customers starting from November 5th.