OpenAI Introduces GPT-5-Codex: A New Version of GPT-5, Optimized Better for Agentic Coding in Codex
The American AI tech brand – OpenAI has now officially introduced its GPT-5-Codex as a new version of GPT-5. It has been optimized for agentic coding in Codex, and has undergone training focusing on real-world software engineering works. Now available on Codex CLI, IDE extension, web, mobile, and also for code reviews in Github, and it is included in ChatGPT Plus, Pro, Business, Edu, and Enterprise plans.
Read more about it below.
OpenAI’s GPT-5-Codex – Officially Introduced
As mentioned, with its training that is based on complex real-world engineering works like developing full projects from scratch, feature additions and tests, debugging, performing large-scale refactors, as well as conducting code reviews, OpenAI has optimized GPT-5-Codex to better handle agentic coding in Codex. Notably, it has been made more steerable, will follow AGENTS.md instructions, and is capable of producing higher-quality codes.
GPT-5-Codex combines two essential skills – First, it pairs with developers in interactive sessions. Second, it offers persistent and independent execution on longer tasks. Also, during its testing stage, GPT-5-Codex was seen working independently for over 7 hours straight on complex and longer tasks. All of its training was done to meet two specific functions – conduct code reviews and find critical flaws. Moreover, OpenAI’s GPT-5-Codex is also claimed to be a reliable partner for front-end tasks as well.
In terms of its availability and pricing, as listed above, the GPT-5-Codex is built for Codex CLI, Codex IDE extension, Codex cloud environment, and Github. Additionally, it also supports versatile tool use. Subscribers of ChatGPT Plus, Pro, Business, Edu, and Enterprise plans will be able to access GPT-5-Codex. Also, for developers who use Codex CLI via API key, OpenAI will be making GPT-5-Codex soon available in the API.
