OpenAI has announced the launch of Codex — a new cloud-based artificial intelligence (AI) agent designed to assist programmers in their work. This tool is capable of writing code, testing it, and creating requests for changes, significantly accelerating the software development process.
This is reported by Business • Media
Codex is already available to users of ChatGPT Pro, Team, and Enterprise versions, and will soon be available for Plus. According to OpenAI representatives, “Codex is a software engineering agent that can work on multiple tasks simultaneously: writing functions, answering questions about code, finding and fixing bugs, and creating requests to include changes in the code.”
Features of Codex
Codex operates in an isolated environment in the cloud, where the user’s codebase is uploaded. It has the ability to read and edit files, run tests, conduct analysis, and perform type checking without needing an internet connection, making its use secure.
This tool is built on the codex-1 model — a specialized version of OpenAI’s o3 technology, which is adapted for programming tasks. Codex was trained using reinforcement learning on real engineering projects to closely match coding styles and team standards.
Implementing Codex in Business
Users can interact with Codex through a sidebar in ChatGPT, delegating tasks and asking about code. Agents log their actions in the terminal log and testing results, ensuring transparency at every stage.
Among the early partners of Codex are companies like Cisco, Temporal, Superhuman, and Kodiak, which are already integrating it into their workflows. Cisco notes: “Codex helps us implement ambitious ideas faster, acting as a smart assistant for engineers.”
Additionally, at Temporal, Codex performs complex tests in the background without distracting developers, while at Superhuman, even product managers can make minor changes to the code without involving an engineer, simply by reviewing them.
OpenAI also introduced a lightweight model, codex-mini-latest, for Codex CLI, which operates in the terminal. It responds more quickly to requests and is better suited for interactive coding. Currently, Codex is available at no additional cost in a research preview format, with plans for a flexible pricing model in the future.
The company plans to implement a rate of $1.50 per 1 million input tokens and $6 per 1 million output tokens, with a 75% discount on repeated requests. OpenAI states: “We envision a future where developers focus on key tasks while delegating the rest to agents. Codex is the first step towards that future.”
Not long ago, OpenAI also launched the OpenAI for Countries partnership initiative, expanding access to the new GPT-4.1 chatbot model for all users.