OpenAI Unveils GPT-5.2-Codex

OpenAI has released GPT-5.2-Codex, a state-of-the-art agentic coding model for complex real-world software engineering and defensive cybersecurity. This model is built on GPT-5.2 and includes the following key improvements.

• Improved long-running task performance through context compaction
• Enhanced performance for large-scale code changes such as refactoring and migrations
• Improved performance and reliability on Windows environments
• Strengthened cybersecurity capabilities

GPT-5.2-Codex achieved top-tier performance on the SWE-Bench Pro and Terminal-Bench 2.0 benchmarks. It can perform complex tasks while maintaining full context within large repositories, and improved visual performance allows it to quickly turn design mockups into functional prototypes.

Its cybersecurity capabilities have also improved dramatically. OpenAI is designing a safe deployment approach based on the Preparedness Framework to manage the dual-use risks that come with the model's strengthened security capabilities. All Codex features are currently available to paying ChatGPT users, and access for API users will be rolled out over the coming weeks.