OpenAI’s GPT-Live-1 API brings full-duplex voice to developers

OpenAI is making GPT-Live-1 available to developers through an API. The speech model supports full-duplex interaction, meaning an app can listen and talk at the same time. The model is already running inside ChatGPT, and developers can pair it with different backend models depending on the task.

Full-duplex speech can make AI conversations feel more natural because the system does not have to wait through rigid turn-taking. Developers can tune deployments around reasoning depth, speed, and cost depending on the use case. The reported price is $0.05 per minute, so teams should weigh the user-experience gains against operating costs.

OpenAI’s benchmarks show GPT-Live-1 ahead of GPT-Realtime-2.1 across several voice-interaction measures. Full-duplex interactivity rises to 80.1 percent from 45.4 percent, while turn-taking latency drops to 0.8 seconds from 1.4 seconds. Tool-calling accuracy improves to 87 percent from 60 percent, and a banking voice support benchmark reaches a 32 percent pass rate, up from 12.4 percent.

GPT-Live-1 ships with twelve new voices covering different accents, dialects, and languages. It also provides ASR transcripts and response text by default. Yelp is using the model for phone-based reservations and reports better call handling, according to CTO Alex Levy.