🔒 GPT Codex: The next coding revolution

OpenAI has released GPT-5-Codex — a special version of the GPT-5 model optimized for programming tasks. The new model combines the speed of interactive sessions with the ability to work independently on large projects for extended periods.

The new version of the model from OpenAI brings a series of improvements that —as the company reports on its blog — will further facilitate developers’ work. GPT-5-Codex can independently write long sections of code, debug, add new features and tests, and thoroughly improve existing code.

Key updates

A key element is intelligent code review. The model can analyze repositories, run tests and point out potential errors before they appear in subsequent project iterations. Importantly, according to tests involving experienced software engineers, comments now generated by GPT-5-Codex are significantly more accurate than those by GPT-5. Inaccurate comments constituted only 4.4% of the total, while in the previous version, this percentage reached 13.7%.

This article is part of the paid edition of hAI Magazine.

Free month for new readers

Activate

Share

You might be interested in