GPT-5.4 Thinking System Card (openai.com)

OpenAI has announced the release of GPT-5.4 Thinking, the latest reasoning model in the GPT-5 series. This new model features an enhanced safety mitigation framework, maintaining continuity with previous iterations but introducing specific safeguards against high-level cybersecurity risks. This is the first general-purpose model to implement these advancements, building on the previous innovations found in GPT-5.3 Codex, ChatGPT, and the API. The introduction of GPT-5.4 Thinking is significant for the AI/ML community as it highlights a proactive stance in addressing cybersecurity challenges with advanced AI systems. By incorporating these specialized mitigations, OpenAI aims to bolster the trustworthiness and accountability of its models in environments where security is paramount. Additionally, developers and organizations utilizing GPT-5.4 Thinking can expect improved protection measures tailored to confront high-capability cyber threats, setting a benchmark for future AI developments in safety and reliability.