OpenAI Codex: a cloud coding agent, one sandbox per task
In one sentence On 16 May 2025 OpenAI presents Codex in ChatGPT as a research preview: a cloud software engineering agent powered by codex-1 (a version of o3 optimized for code) that works on several tasks in parallel, each in a sandbox loaded with the user's repository.
On 16 May 2025 OpenAI reused the Codex name for a new product: a programming agent that works in the cloud, inside ChatGPT.
You connect your repository and assign tasks: write a function, fix a bug, answer questions about the code, propose a pull request. Each task runs in an isolated cloud environment with a copy of the repository, and several tasks can run in parallel. The agent can run the tests and repeat them until they pass, then shows the result for review.
The model is codex-1, a version of o3 trained with reinforcement learning on real-world coding tasks. At launch it is a research preview for ChatGPT Pro, Enterprise and business-plan subscribers, with Plus and Edu to follow.
Companies
OpenAI
Tools
OpenAI Codex, codex-1, ChatGPT
Tags
Sources