OpenAI has expanded its GPT-Live voice model to its coding platform Codex and the ChatGPT desktop app, allowing developers to write and manage code using only their voice. The feature, announced recently, lets engineers speak commands to perform tasks like reviewing code, debugging, or starting new work without typing or clicking.
GPT-Live is a voice model that can listen and speak at the same time, making conversations feel more natural. It was first launched in July 2026. Now, it works inside the ChatGPT desktop app on Mac and Windows. Developers can talk to the system while it runs complex tasks in the background using other AI models.
With this update, a developer can ask the system to check a bug, review a change, and write tests all at once using just one spoken request. The tool can also look at open windows, local files, and connected services like GitHub. OpenAI says this is the first time voice control has been built directly into Codex on the desktop.
The feature is available only to paying users of Plus, Pro, Business, Enterprise, or Education plans. It uses the same usage limits as other tasks in Codex and ChatGPT Work.
Codex has more than 5 million weekly active users. Some developers have already expressed excitement about the hands-free approach, saying it brings the technology closer to personal AI assistance.