OpenAI launches GPT-5 Codex for software coding

An updated version of GPT-5 called as GPT-5 Codex optimized for practical software engineering, has been formally released by OpenAI. The release is a significant step toward establishing AI as an active coding collaborator, able to work independently on lengthy, intricate projects or in tandem with developers.

Unlike earlier iterations, GPT-5 Codex is designed not just to generate code but to build full projects, add features, debug, conduct large-scale refactors, and perform code reviews. Its adaptive reasoning allows it to respond quickly for small, well-defined requests while dedicating hours of iterative effort to larger tasks. During internal tests, the model successfully worked independently for over seven hours, fixing failures and delivering working implementations.

A major highlight is its code review capability, which OpenAI says identifies critical flaws with greater accuracy. In evaluations against open-source commits, GPT-5 Codex’s review comments proved more relevant and reliable, ensuring developers’ attention is directed toward issues that matter most.

GPT-5 Codex comes bundled with ChatGPT Plus, Pro, Business, Edu, and Enterprise plans, with usage scaling based on plan type. Enterprise customers also benefit from shared credit pools for team-wide flexibility. The tool will be accessible over CLI, IDE extensions, GitHub, the web, and even the ChatGPT iOS app. Developers can seamlessly switch between local and cloud environments without losing context. 

To ensure safety, Codex operates in sandboxed environments with network restrictions, asks permission before executing sensitive commands, and provides logs and test results for transparency. OpenAI emphasizes Codex should act as an additional reviewer, not a replacement for human oversight.

With GPT-5 Codex, OpenAI moves closer to its vision of AI as a trusted coding partner—a teammate that adapts to context, enhances productivity, and strengthens reliability in modern software development.