OpenAI opens GPT-Live-1 full-duplex voice model to developers at $0.05 per minute

OpenAI released GPT-Live-1 in its API on September 10, giving developers access to the full-duplex speech model already running inside ChatGPT. Full-duplex means the model can listen and talk at the same time, responding to interruptions and acknowledgements as they happen. Traditional voice agents chain speech-to-text, a reasoning model, and text-to-speech, and each handoff adds latency and risks losing timing or context. GPT-Live-1 handles listening and speaking in a single model that reasons over incoming and outgoing audio together, while delegating deeper reasoning and actions to whatever backend models and tools the developer pairs it with. Developers can match reasoning depth, speed, and cost to each use case. The front-end voice layer costs $0.05 per minute, which The Decoder describes as not cheap. On OpenAI's own benchmarks the model scores 80.1 percent on full-duplex interactivity tests compared with 45.4 percent for GPT-Realtime-2.1, cuts turn-taking latency from 1.4 seconds to 0.8 seconds, and raises tool-calling accuracy from 60 to 87 percent. In a banking voice support benchmark the pass rate rises from 12.4 to 32 percent. Yelp is using the model for phone-based reservations and reports better call handling, according to its CTO Alex Levy. The release includes twelve new voices across accents, dialects, and languages, and provides transcripts and response text out of the box.