OpenAI Launches Codex, a Powerful AI Agent for Coding

OpenAI revealed a research preview of Codex, a new AI coding agent designed to transform software engineering.

Built on the codex-1 model (a specialized version of OpenAI’s o3 reasoning model), Codex can write code, fix bugs, run tests, and answer questions about a customer’s codebase—all autonomously in the cloud.

Codex will be available to ChatGPT Pro, Team, and Enterprise users, aiming to help developers delegate more complex tasks and streamline workflows.

Unlike many current tools that assist developers in real time, Codex works independently and delivers results within minutes, adapting to an organization’s coding style and aiding in code reviews.

OpenAI’s Vice President of Engineering, Srinivas Narayanan, described Codex as “a fundamentally new way of working” and highlighted the company’s plan to gather user feedback during this early research phase.

While OpenAI leads consumer-facing AI chatbots, the coding AI space is competitive, with rivals like Anthropic’s Sonnet models gaining popularity among developers.

OpenAI is reportedly considering acquiring AI coding startup Windsurf for $3 billion, signaling a major investment in this fast-growing domain.