OpenAI has officially launched Codex, its latest AI-powered coding assistant integrated into ChatGPT. Described as a cloud-based software engineering agent, Codex can handle multiple software development tasks simultaneously.
Powered by codex-1, a variant of OpenAI’s o3 AI reasoning model, Codex is designed to streamline workflows and boost developer productivity. It helps with everything from writing features and fixing bugs to managing pull requests and answering technical questions—all in a secure cloud-based sandbox connected to your codebase.
On Friday, OpenAI introduced its new AI coding agent called Codex during a live stream. “It is a coding AI agent built into ChatGPT and is capable of managing several development tasks at once.” OpenAI highlights that Codex can “generate features, fix bugs, and answer technical questions about a user’s codebase,” with every task securely executed in its own isolated environment.
Codex runs on a custom-tuned version of OpenAI’s o3 reasoning model. “This model was trained through reinforcement learning on real-world development scenarios to produce code that ‘closely mirrors human style and PR preferences, adheres precisely to instructions, and can iteratively run tests until it receives a passing result’.”
Codex is now being gradually made available to subscribers of ChatGPT Pro, Enterprise, and Team plans. This staged release ensures priority access to advanced AI coding tools for high-tier users.
“Support for users on the ChatGPT Plus and Edu plans is expected to follow shortly,” although OpenAI has not announced an exact timeline yet.
To access Codex, users can find it in the ChatGPT sidebar. Just type a prompt and choose “Code” to assign tasks, or click “Ask” to seek help with understanding your codebase. It’s a seamless way to engage the AI for coding-related queries and automation.
Depending on the task’s complexity, Codex completes jobs in 1 to 30 minutes, all while you watch the workflow unfold in real time. “Each task is handled separately in its own isolated workspace,” and Codex has permission to read, edit, and run your codebase using test suites, linters, and other tools.
Once Codex finishes a task, it commits changes within its controlled environment. “It offers transparent, verifiable proof of its actions by referencing terminal logs and test results.” From there, users can review the outcome, request further changes, initiate a GitHub pull request, or apply changes directly to their development system. You can also configure Codex’s environment to replicate your local setup.
Codex marks a new era in AI-assisted software development, blending automation, intelligence, and human-like reasoning into a single tool within ChatGPT. With its ability to manage multiple tasks—from bug fixes to generating production-ready code—Codex empowers developers to work faster and more efficiently.
OpenAI’s integration of the o3 model means Codex doesn’t just follow commands but understands and refines code through testing. Its real-time feedback, sandboxed execution, and verifiable task logs promote transparency and control for developers.
The initial rollout to Pro and Enterprise users also reflects OpenAI's strategy to deliver innovation to its most engaged user base first. As broader access approaches, Codex is poised to revolutionize workflows for students, educators, and professionals alike.
With a future pointing towards more autonomous, intelligent software agents, Codex demonstrates OpenAI’s ambition to make coding more accessible and dynamic for all. It’s not just an assistant—it’s a powerful collaborator in your dev environment.