GPT-Live-1 Voice Model Now Available via API

On September 10, 2026, OpenAI opened the API for GPT-Live-1, a full-duplex voice model that listens and speaks simultaneously, replacing the STT–LLM–TTS cascade with a single speech stack while delegating reasoning and tool calls to a backend model.

Compiled by Sergey KostenchukPublished 2026-09-11Updated 2026-09-18

2026-09-11 Coding OpenAI

🛠 GPT-Live-1 Voice Model Now Available via API

On September 10, OpenAI opened the API for GPT-Live-1 — a full-duplex voice model from ChatGPT's voice mode. It listens and speaks simultaneously: a single speech stack replaces the STT–LLM–TTS cascade, and interruptions and background noise are handled within a single model. Reasoning and tools are delegated to a backend model — the conversation continues while it computes. The frontend costs $0.05 per minute.

🌍 Voice agents are no longer a pipeline of 'recognition + LLM + synthesis': the voice line does not drop while the backend executes tools, and it can be replaced with a third-party model.

👤 Developers connect GPT-Live-1 via API: 12 new voices, ASR transcripts, telephony, and tempo and tone settings. Users of voice bots will be able to interrupt the agent without pauses while it thinks.

Source 1: https://openai.com/index/introducing-gpt-live-1-in-the-api/ Source 2: https://developers.openai.com/api/docs/guides/live

Sources