OpenAI launches GPT-Live-1 for full-duplex voice agents
OpenAI is putting its natural, interruptible voice layer directly into the API at $0.05 per minute.
GPT-Live-1 processes incoming and outgoing audio together instead of chaining transcription, reasoning, and speech synthesis. The piece says that should reduce latency, missed timing, and brittle handoffs in voice agents. Developers can steer voice style, pair it with separate backend models and tools, and use it for telephony workflows like reservations, order updates, and support. Early users cited fewer interruptions and leaner voice-agent codebases. TestingCatalog's note
GPT-Live-1 processes incoming and outgoing audio together instead of chaining transcription, reasoning, and speech synthesis. The piece says that should reduce latency, missed timing, and brittle handoffs in voice agents. Developers can steer voice style, pair it with separate backend models and tools, and use it for telephony workflows like reservations, order updates, and support. Early users cited fewer interruptions and leaner voice-agent codebases. TestingCatalog's note
score 7