OpenAI Unveils GPT-5-Codex: A Game Changer in Agentic Software Engineering
OpenAI has rolled out a significant update to Codex, its AI coding agent. The update introduces GPT-5-Codex, a new version of OpenAI's latest GPT-5 model tailored for agentic software engineering. The update brings several enhancements, including the ability to attach and share images with the coding agent.
The new GPT-5-Codex has shown impressive performance in refactoring code and conducting code reviews, with a low error rate and high-impact comments. It uses dynamic reasoning to tackle complex problems and can work autonomously for extended periods. The update includes a rebuilt Codex CLI centered around agentic workflows, a new IDE extension for tools like VS Code, and integration with GitHub for code reviews. The model uses an AGENT.md file to guide its coding guidelines and other instructions. Codex now uses a to-do list to track progress and has improved tool calls and user interface for following along with the model's actions.
The update also includes improvements to the Codex web service, with a 90% reduction in median completion time for new tasks and follow-ups. Access to Codex is included for OpenAI users with ChatGPT Plus, Pro, Business, Edu, and Enterprise plans, with usage limits scaling based on the plan.
The launch of GPT-5-Codex brings significant improvements to OpenAI's Codex, enhancing its capabilities in agentic software engineering. With faster task completion times and new features like image sharing, developers can expect a more efficient and intuitive coding experience.
