OpenAI's new software-driven model, GPT-5-Codex

OpenAI aims to boost its growth with a new AI model. Designed for those focused on software development, the new GPT-5 Codex will allow users to write, edit, and perform other tasks related to code.

GPT-5-Codex, a specially optimized version of the GPT-5 model, is a new version of the coding tool Codex.

This model was developed to provide higher performance on the software engineering side.

The new model allows for tasks such as creating projects from scratch, adding features to existing projects, or adding tests. GPT-5-Codex achieved a SWE-bench Verified score of 74.5, surpassing GPT-5 High's score of 72.8.

In the code reconstruction task, GPT-5-Codex scored 51.3 percent, while GPT-5 got 33.9.

The new model's ability to decide how much thought it needs to do depends on the complexity of the task is noteworthy. It was reportedly able to work autonomously for over seven hours on large-scale, complex tasks.

It is reported that GPT-5-Codex is available for code reviews on web, mobile and GitHub, as well as the Codex CLI and IDE plugin, as part of the ChatGPT Plus, Pro, Business, Edu and Enterprise plans.