GPT-Live-1

OpenAI's full-duplex voice models (GPT-Live-1 and GPT-Live-1 mini): simultaneous listening and speaking — the model can be interrupted mid-sentence and can interject — with live translation, rolling out globally in ChatGPT. Complex queries delegate to GPT-5.5 behind the scenes. Proprietary; system card published July 8. The developer-facing siblings (gpt-realtime-2.1 / -2.1-mini) shipped July 6 with reasoning and tool use at the mini tier and ~25% lower p95 latency.

GPT-Live-1 reached general availability on the API on 10 September 2026 as gpt-live-1: full-duplex, billed at $0.05 per minute with the backend model billed separately; Artificial Analysis scores it 81.5 on its Speech-to-Speech Index, second to Gemini 3.8 Live Extended Thinking.