OpenAI's GPT-Live-1 Processes Speech Both Ways at Once, Ditching Turn-Based Delays

OpenAI released GPT-Live-1, a full-duplex voice model that handles incoming and outgoing audio simultaneously, eliminating wait times between exchanges. Available now on iOS, Android, and web, it manages mid-sentence interruptions, real-time translation, and natural verbal cues like backchannels. Complex queries route to GPT-5.5. Paid subscribers get full access; free users receive GPT-Live-1 mini.