OpenAI launches GPT-Live, a full-duplex voice upgrade that lets ChatGPT talk more like a person

OpenAI has launched GPT-Live, a pair of new voice models that replace the company's existing Advanced Voice Mode in ChatGPT. The defining change is architectural: rather than waiting for a user to finish speaking before generating a reply, GPT-Live is full-duplex, meaning it can listen and speak at the same time. VentureBeat reports the launch came Wednesday, and OpenAI frames it as a fundamental redesign of how people talk to ChatGPT rather than an incremental tuning of the old system.

The significance of full-duplex operation is that it removes the rigid turn-taking that has defined voice assistants until now. Human conversation is full of overlaps, interruptions, and mid-sentence adjustments, and a system that must wait for silence before responding always feels like a walkie-talkie rather than a dialogue partner. By handling listening and speaking simultaneously, GPT-Live can in principle be interrupted naturally, react while the user is still talking, and avoid the awkward pauses that made earlier voice modes feel mechanical.

Shipping this as a replacement for Advanced Voice Mode, rather than an optional extra, signals that OpenAI sees natural spoken interaction as a core interface for ChatGPT going forward. Voice is one of the areas where model quality alone does not decide the experience; latency and conversational flow matter just as much, and those are exactly the properties a full-duplex architecture targets. Competitors building their own real-time voice systems will now be measured against an assistant that no longer waits its turn to speak.