OpenAI's GPT-Live can now listen and talk at the same time

OpenAI shipped GPT-Live, a new voice model family that ditches the old take-turns loop for a full-duplex architecture — it listens and speaks simultaneously, deciding many times per second whether to talk, pause, interrupt, or fire off a tool call.

Two flavors are rolling out globally across iOS, Android, and the web: GPT-Live-1 as the default for paid users and GPT-Live-1 mini for the free tier. For anything needing web search or heavy reasoning, it hands off to a frontier model in the background — currently GPT-5.5 — then folds the answer back into the chat. It also drops human-ish cues like "mhmm" and "got it" so you know it's tracking. API access is coming.

The timing isn't random: it lands days before OpenAI's broader GPT-5.6 launch (Sol, Terra, and Luna variants), with preview access already widening.

Why it matters: real-time, interruptible voice is the interface shift that makes talking to AI feel less like a walkie-talkie and more like a conversation.