Last month, OpenAI released the GPT-5.2 model series, delivering improved intelligence in ChatGPT for general users and through its API for developers. The GPT-5.2 models set new state-of-the-art scores across several AI benchmarks.
Coding remains one of the most popular use cases for AI models, and OpenAI has been releasing Codex-specific models tailored for developers. For example, OpenAI introduced GPT-5.1-Codex, a version of GPT-5.1 optimized for agentic coding tasks in Codex and other developer environments. This was followed by GPT-5.1-Codex-Max, a frontier-class agentic coding model designed for long-running and complex tasks.
Last month, OpenAI announced GPT-5.2-Codex, its most advanced agentic coding model to date. The model is optimized for agentic workflows in Codex and includes improvements such as better long-horizon performance through context compaction, stronger handling of large codebases, and overall higher reliability. Following positive feedback from Codex developers, OpenAI has now announced that GPT-5.2-Codex is available via its API for developers.
GPT-5.2-Codex is now available in the Responses API—the same model available in Codex.
— OpenAI Developers (@OpenAIDevs) January 14, 2026
It’s strong at complex long-running tasks like building new features, refactoring code, and finding bugs.
Plus, it’s the most cyber-capable model yet, helping to find and understand codebase… https://t.co/oN2lAsrV4D
GPT-5.2-Codex is already supported in popular IDEs, including Cursor, Windsurf, Factory, GitHub, and others. The model achieves state-of-the-art performance on benchmarks such as SWE-Bench Pro and Terminal-Bench 2.0.
Cursor tweeted on X that it believes GPT-5.2-Codex is a frontier model for long-running tasks. Using this model through Cursor, the team was able to build a web browser entirely from scratch. Cursor’s CEO revealed that the model ran uninterrupted for an entire week and ultimately produced more than three million lines of code across thousands of files. The resulting browser includes a from-scratch rendering engine written in Rust, featuring HTML parsing, CSS cascading and layout, text shaping, painting, and a custom JavaScript virtual machine.
We are only at the beginning of 2026. By the end of the year, we can expect even more capable coding models from OpenAI, Anthropic, and Google, models that could realistically replace entry- and mid-level software engineers for many modern software development tasks.