OpenAI has introduced GPT-Live, a new family of voice models designed to make conversations with ChatGPT feel more like talking to a real person. The system now powers ChatGPT Voice on iOS, Android, and the web, and has been available worldwide since launch.
Full Duplex Instead of Waiting for Silence
The core feature of GPT-Live is full-duplex communication: the model can listen and respond at the same time, without waiting for the other person to stop talking. In practice, this means smoother turn-taking, small verbal cues, and natural thinking pauses, without the noticeable lag of older voice systems.
Earlier versions of ChatGPT Voice relied on three chained models: one for transcription, one for generating a response, and one for converting that response back to audio. The Advanced Voice Mode improved on this with a single model, but it still used silence to detect the end of a turn. Background noise or a brief pause could cause it to cut in at the wrong moment. GPT-Live processes audio continuously, deciding several times per second whether to speak, stay quiet, or hand off a task.
Two Variants and API Plans
GPT-Live comes in two versions. GPT-Live-1 is the default for Go, Plus, and Pro subscribers. GPT-Live-1 Mini serves free-tier users. For tasks that require search or deeper reasoning, the model can route to GPT-5.5 in the background without interrupting the conversation. An API is planned, according to OpenAI.
Internal evaluations using human testers showed a preference for GPT-Live over the Advanced Voice Mode, particularly in conversational flow and turn transitions. These figures come from OpenAI's own tests and have not been independently verified.