OpenAI Introduces GPT-5.5-Cyber for High-Impact Cybersecurity Research

OpenAI released GPT-5.5-Cyber, a version of its GPT-5.5 model optimized specifically for cybersecurity research, available through the company's Trusted Access for Cyber (TAC) program. The model achieved an 81.9% score on the CyberGym benchmark comprising over 1,500 historical vulnerabilities. Unlike earlier TAC models that could only generate exploitation plans, GPT-5.5-Cyber can validate exploits by launching simulated attacks, making it useful for automated red teaming. OpenAI is initially limiting access to defenders securing critical infrastructure and has implemented enhanced misuse monitoring. The model competes directly with Anthropic's Claude Mythos Preview for vulnerability research.