OpenAI has unveiled GPT-5.3-Codex, a new version of its Codex model that expands beyond coding assistance to function as a computer-using agent capable of handling a wider range of professional tasks.
The company describes the release as a significant upgrade from GPT-5.2-Codex, combining strong coding performance with enhanced reasoning and deeper professional knowledge in a single system. OpenAI says the model runs 25 percent faster than its predecessor, enabling quicker task execution and more responsive interactions.
GPT-5.3-Codex is now available to users on paid ChatGPT plans wherever Codex is supported, including the Codex app, command line interface, integrated development environment extensions, and the web version of ChatGPT. OpenAI confirmed it is working toward enabling API access in the future.
One of the notable changes is how the model handles everyday website-related requests. OpenAI said that when users provide underspecified prompts, GPT-5.3-Codex generates sites with more practical functionality and more sensible default configurations than earlier versions.
The company also emphasized improvements in how users supervise long-running tasks. The Codex app now provides frequent updates on key decisions and progress, allowing users to step in, guide the system, or adjust direction while work is still underway.
According to OpenAI, GPT-5.3-Codex can incorporate feedback mid-task and maintain context throughout extended interactions. The experience, the company said, is designed to feel less like issuing one-off commands and more like steering a knowledgeable colleague through complex work.
With GPT-5.3-Codex, OpenAI is positioning Codex as not just a coding assistant, but a broader AI agent built to support professional workflows across multiple domains.
News Source: Emirates News Agency
