OpenAI Releases GPT-Live-1 API for Full-Duplex Real-Time Voice
Original title:OpenAI's GPT-Live-1 API lets developers build apps that talk and listen at the same time
OpenAI has released GPT-Live-1 as a developer API, bringing full-duplex conversational audio to third-party applications. The model processes incoming user speech while simultaneously streaming voice output, scoring 80.1 percent on interactivity benchmarks compared to 45.4 percent achieved by its predecessor. Priced at $0.05 per minute, the API addresses the friction of rigid turn-taking in voice agents, while establishing a clear commercial rate for continuous conversational audio.
Why it's worth reading
Full-duplex audio overcomes the awkward turn-taking delays of previous voice assistants; an 80.1 percent interactivity score alongside transparent per-minute pricing provides a concrete baseline for real-time speech products.