OpenAI’s GPT-5.4-Cyber is not for everyone, just like Anthropic’s Claude Mythos: Here’s how it compares

Just weeks after Anthropic’s headline-grabbing announcement on Claude Mythos – the dangerously powerful AI model that’s been restricted to a couple of infrastructure organisations – arch rival OpenAI has unveiled GPT-5.4-Cyber. This is a specialised cybersecurity-focused variant of its latest GPT-5.4 model that has been restricted to a carefully controlled, limited rollout.