OpenAI has launched Codex, a powerful new cloud-based AI software engineering agent now integrated into ChatGPT.
As per OpenAI, Codex, designed to handle a wide range of development tasks simultaneously, Codex can write new features, fix bugs, and answer questions about a user’s codebase — all within secure, isolated environments called sandboxes.
Codex AI: A Smart Agent for Developers
According to OpenAI, Codex is capable of:
- Writing functional code based on prompts
- Fixing bugs in existing code
- Explaining complex code or systems
- Running unit tests, linters, and type checkers
Each task runs in its own private sandbox, ensuring safety and task isolation while maintaining performance.
Powered by OpenAI’s Latest Model
Codex is powered by a special version of OpenAI’s o3 model which was optimized for software engineering related task. This model was trained using reinforcement learning on real-world coding environments, enabling it to “closely mirrors human style and PR preferences, adheres precisely to instructions, and can iteratively run tests until it receives a passing result”.:
- Write code that mimics real human coding styles
- Follow precise instructions
- Run and validate tests until the code passes
OpenAI says Codex is built to match the pull request (PR) style used by developers in professional workflows, making its output feel natural and production-ready.
OpenAI SEO Sam Altman shared his excitement on X (formerly Twitter), stating:
“It is amazing and exciting how much software one person is going to be able to create with tools like this.”
Availability and Access of Codex AI
Codex is now rolling out to ChatGPT Pro, Team, and Enterprise users. OpenAI also plans to expand access to Plus and Edu users in the coming weeks.
To use Codex:
- Click on the “Code” button in the ChatGPT sidebar to assign a new task
- Use the “Ask” button to query Codex about your codebase
OpenAI in there post stated, “Codex can read and edit files, as well as run commands including test harnesses, linters, and type checkers.”
Real-Time Task Execution and Transparency
Each task handled by Codex typically takes between 1 and 30 minutes, depending on complexity. Users can monitor progress in real time.
OpenAI stated, “Task completion typically takes between 1 and 30 minutes, depending on complexity, and you can monitor Codex’s progress in real time.”
To ensure transparency, Codex provides verifiable logs and test outputs as evidence for its actions. This helps users trace every step and confirm that the AI did not hallucinate or make up results.
OpenAI Releases Codex as a Research Preview with Focus on Safety and Transparency
OpenAI has launched Codex in research preview mode, following its careful, step-by-step rollout strategy.
They stated, “We’re releasing Codex as a research preview, in line with our iterative deployment strategy. We prioritized security and transparency when designing Codex so users can verify its outputs – a safeguard that grows increasingly more important as AI models handle more complex coding tasks independently and safety considerations evolve. Users can check Codex’s work through citations, terminal logs and test results. When uncertain or faced with test failures, the Codex agent explicitly communicates these issues, enabling users to make informed decisions about how to proceed. It still remains essential for users to manually review and validate all agent-generated code before integration and execution.”
