🔎 OpenAI releases GPT-5.4-Cyber to select users, challenging Anthropic’s Mythos GPT-5.4-Cyber will be available to some participants in OpenAI Trusted Access for Cyber program: • The new model is aimed at finding issues in software to help organizations fix them. • It places fewer constraints on how users can probe the model for cybersecurity issues, OpenAI said. • No specific benchmarks are available to assess how the model differs from the regular GPT-5.4, suggesting GPT-5.4-Cyber might be worse than Claude Mythos. #OpenAI #Anthropic #Mythos