OpenAI has unveiled Codex, its most advanced AI-powered coding assistant yet, marking a significant leap in the evolution of AI tools designed for software development. Currently in research preview, Codex is available to ChatGPT Pro, Team, and Enterprise users, with plans to extend access to ChatGPT Plus and Edu subscribers in the near future.

Codex: A Specialized Engineering Agent

Built on Codex-1, a customized version of OpenAI’s o3 reasoning model, Codex is tailored specifically for software engineering. It goes beyond traditional autocomplete or code suggestion tools by autonomously writing, testing, and debugging code until successful results are achieved. The tool is designed to follow instructions with high fidelity and generate more accurate, production-grade code.

Unlike simple code completion assistants, Codex operates as a cloud-based agent capable of performing tasks in parallel and sustaining its work over long durations—up to 30 minutes at a time—without needing constant user input or access to the user’s device or browser.

Integrated Development Capabilities

Codex works within a secure, isolated virtual environment and integrates directly with GitHub, allowing it to access, analyze, and modify entire codebases. It can add new features, fix bugs, generate pull requests, and respond to technical queries, functioning like a “virtual coworker”, according to Josh Tobin, OpenAI’s Agents Research Lead.

Users can interact with Codex directly inside ChatGPT by choosing the “Code” option or clicking the “Ask” button for specific programming-related questions. All tasks initiated by Codex—active or completed—are displayed beneath the prompt area, offering full visibility into its decision-making and progress.

New Codex CLI and Pricing Details

Alongside the main Codex platform, OpenAI is updating its Codex Command Line Interface (CLI), an open-source AI agent developers can use in their local terminals. The latest CLI version leverages the o4-mini model, optimized specifically for programming tasks. It is available via the OpenAI API at a rate of $1.50 per million input tokens and $6 per million output tokens.

Security and Responsible AI Use

OpenAI emphasizes the safety and security of Codex. It does not generate malicious or harmful code, and its execution environment is offline and sandboxed, minimizing the risk of misuse. Nevertheless, OpenAI acknowledges the limitations of current coding AI, citing a Microsoft study that highlights persistent challenges in debugging and verification.

Competitive Landscape and Strategic Moves

Codex enters a rapidly growing market for AI developer tools. Competitors include Anthropic’s Claude Code, Google’s Gemini Code Assist, and rising platforms such as Cursor and Windsurf. Cursor, for example, is reportedly generating $300 million in annual revenue and seeking a $9 billion valuation.

Amid this competitive surge, rumors suggest OpenAI has acquired Windsurf for $3 billion, signaling its aggressive expansion into the AI software development space.

Part of a Larger AI Ecosystem

Codex is a cornerstone in OpenAI’s broader vision of transforming ChatGPT into a comprehensive AI platform. Recent additions include Sora for video generation, Deep Research, and Operator for web browsing—tools that together aim to provide a seamless, multi-functional AI workspace for premium ChatGPT users.

Redefining Software Engineering

By enabling AI agents to function as autonomous collaborators, OpenAI’s Codex is reshaping what’s possible in software development. It not only assists but actively contributes to complex engineering workflows, paving the way for a future where AI is a full-fledged partner in the coding process.