OpenAI's GPT-5 Codex: ChatGPT parent upgrades AI coding tool

ChatGPT parent OpenAI has optimised its latest artificial intelligence (AI) model, GPT-5, for its coding agent, Codex. The AI major said the new agentic software development tool can handle long, complex tasks equally well, as quick, interactive sessions.

Announcing the latest version of the AI coding tool, OpenAI CEO Sam Altman said it already accounts for 40% of the traffic on Codex.



GPT-5-Codex: Key features

OpenAI claims GPT-5-Codex has been trained to build full projects from the ground up, add features and tests, debug code, improve the internal structure of code on a large scale (code refactoring), and conduct code reviews.

GPT-5-Codex dynamically adapts the time it spends on "thinking" about a task depending on its complexity, OpenAI said. It can pair with developers in small interactive sessions, as well as work independently on longer tasks. This would yield snappier responses from Codex on small, well-defined requests, and it can work longer on complex tasks such as big refactors, the company added.

OpenAI claimed GPT-5-Codex worked independently for more than seven hours at a time on large, complex tasks, iterating on its implementation, fixing test failures, and ultimately delivering a successful implementation.

The company added that GPT-5-Codex has been "trained specifically for conducting code reviews and finding critical flaws." It can catch critical bugs in the code before they ship, it added. Reviews by experienced software engineers showed "comments by GPT‑5-Codex are less likely to be incorrect or unimportant, reserving more user attention for critical issues".

GPT-5-Codex: Availability and pricing

OpenAI said GPT-5-Codex is available through a terminal or an integrated development environment (IDE), on the web, and Github. It is also available with ChatGPT Plus, Pro, Business, Edu, and Enterprise plans.

GPT-5-Codex vs Google Gemini Anthropic Claude

OpenAI launched Codex CLI in April and the research preview of Codex in May to claim a portion of the burgeoning AI coding market. In this space, it competes with Microsoft-owned GitHub, Google's Gemini and Anthropic's Claude, as well as startups such as Anysphere and Windsurf.

The company launched GPT-5 last month to a tepid response. The AI major had touted the model as “our strongest coding model to date”, capable of building a game from a single prompt.