OpenAI Launches GPT-Live-1 in the API: Full-Duplex Voice for Enterprises

It is a new OpenAI voice model available in the API that supports full-duplex conversations instead of turn-taking, with stronger instruction following, custom voices, and telephony integration.

Earlier models worked on a listen-then-respond basis. GPT-Live-1 can speak and listen simultaneously, making interactions feel more human, cutting perceived latency, and allowing natural interruption.

Technically yes via the API, but teams should start with bounded use cases such as contact centers or government services, test Arabic dialect quality, and review local data-protection compliance first.

Contact centers, government voice services, banking support, telecom and retail customer care, plus internal employee voice assistants.