GPT-Live-1: OpenAI Launches a Full-Duplex Voice API
OpenAI has released GPT-Live-1 — a voice API that listens and speaks at the same time. Interactivity hit 80.1% vs. 45.4% for its predecessor, latency is 0.8 s, tool-call accuracy is 87%. Priced at $0.05 per minute; Yelp is already using it for reservations.
OpenAI has released GPT-Live-1 as a public API — a full-duplex voice model that listens and speaks simultaneously, without waiting for a pause. Interactivity jumped from 45.4% to 80.1%, latency dropped from 1.4 to 0.8 seconds, and tool-call accuracy improved from 60% to 87%. Pricing is $0.05 per minute; Yelp has already deployed the model for phone reservations.
Voice agents are no longer question-and-answer systems: the duplex model holds a real conversation, interrupts, and reacts in real time. Developers get ASR, response text, and support for 12 new voices — without having to build a pipeline from scratch.
Source: the-decoder.com
Free course
Stop reading about AI — start building with it
The free Claude Code course: your first site, tool or game — no coding. No upsells, no cross-sells — nothing to buy here.
Start free →
Author
Evgenii Arsentev
PhD · Chief Executive Officer, digital health
Articles · Latest articles