OpenAI GPT 5.5 Cyber: Outperforms Mythos 5 in Benchmarks

OpenAI has updated its GPT 5.5 Cyber AI model, claiming it outperforms Mythos 5 on an important benchmark. This new version is designed for cybersecurity work and is described as OpenAI's strongest model yet for finding and patching software vulnerabilities. What's interesting is that the updated GPT 5.5 Cyber scored 85.6% on CyberGym, a benchmark that measures an AI's ability to reproduce known software vulnerabilities. The standard GPT 5.5 model scored 81.8% on the same test. The new model also significantly improved on ExploitGym, achieving 39.5% compared to 25.95% for GPT 5.5. The company states this model is "more permissive" and "more capable for advanced, authorised cybersecurity work." It can analyze large codebases, find security components, check for vulnerable code, and test fixes. OpenAI intends GPT 5.5 Cyber for verified defenders needing advanced capabilities. This development could mean stronger tools for cybersecurity professionals.