Summary
OpenAI introduced GPT-Live, a new generation of voice models built on a full-duplex architecture that lets the system listen and speak simultaneously for more natural back-and-forth conversation. Two versions, GPT-Live-1 and GPT-Live-1 mini, began rolling out to ChatGPT users globally on July 8, 2026, with API access planned. GPT-Live uses GPT-5.5 in the background and delegates to a frontier model for search or deeper reasoning.
What changed
OpenAI shipped GPT-Live-1 and GPT-Live-1 mini, full-duplex voice models powering ChatGPT Voice, with an API release announced as forthcoming.
Why it matters
Full-duplex voice moves conversational AI past rigid turn-taking toward interaction that feels continuous, a step change for voice agents and assistants. The delegate-to-frontier-model design points to a layered architecture where a fast voice front end hands off to heavier reasoning, a pattern other agent platforms will likely adopt.
Evidence excerpt
GPT-Live is built on a full-duplex architecture, meaning it can listen and speak at the same time; at launch it uses GPT-5.5 in the background and delegates to a frontier model for deeper work.