OpenAI has recently made an announcement that its most cutting-edge agentic programming model, GPT-5.2-Codex, is now officially accessible to developers through an API. Initially introduced within OpenAI's proprietary Codex development environment, this model is now being made available to a wider spectrum of developers via the Responses API. Its primary focus is on tackling complex, long-haul software development tasks.

Built upon the GPT-5.2 architecture, GPT-5.2-Codex boasts features such as long-context compression, extensive code refactoring capabilities, and improved performance within Windows environments. In benchmark tests, it achieved an impressive 56.4% accuracy on SWE-Bench Pro and a notable 64% on Terminal-Bench 2.0. For paid ChatGPT subscribers, the model is already available for use across all Codex interfaces, with API access slated to commence in the upcoming weeks.