GPT-Live-1 brings full-duplex voice agents to OpenAI’s API

OpenAI has released GPT-Live-1 in the API for full-duplex voice conversations. The model can listen and speak at the same time, handle interruptions, and delegate deeper reasoning or tool calls to a backend model or agent.

Voice that can handle interruptions

GPT-Live-1 is designed to reason over incoming and outgoing audio together rather than relying only on a chained speech-to-text, language-model, and text-to-speech pipeline. OpenAI says that helps the model respond more naturally when a speaker pauses, changes direction, or talks over the response.

Developers can steer an agent’s tone, pace, and conversational style through the system prompt. The model also handles background noise and silence without narrating every internal step, and OpenAI positions it for longer conversations where context retention matters.

Delegate the hard work

GPT-Live-1 can delegate reasoning and tool use to a backend text model such as GPT-6 Astra or to a third-party model. That lets the voice layer manage the conversation while a separate model or agent handles a complex lookup, a transaction, or another tool-backed action.

OpenAI describes use cases including phone-based reservations and customer support. The API provides audio and text input and output, streaming, and function calling. Developers should treat any tool invocation as a normal high-impact agent action: authenticate it, log it, and require confirmation where a call could change an account or create an obligation.

Availability and pricing

The model identifier is gpt-live-1. OpenAI’s documentation lists the voice layer at $0.05 per minute, billed by the second; backend model and tool usage is charged separately. The API documentation lists concurrent-session limits by usage tier and says the free tier is not supported.

GPT-Live-1 does not support image or video input, structured outputs, fine-tuning, or predicted outputs. Teams evaluating it should test interruption behavior, background-noise handling, latency, and the handoff between voice and backend reasoning on their own calls.

Sources

• OpenAI — “Build more natural voice experiences with GPT-Live-1 in the API” (September 10, 2026): https://openai.com/index/introducing-gpt-live-1-in-the-api/
• OpenAI Developers — “GPT-Live-1” model documentation: https://developers.openai.com/api/docs/models/gpt-live-1
• OpenAI Developers — API changelog: https://developers.openai.com/api/docs/changelog
• OpenAI Developers on X — official developer updates: https://x.com/OpenAIDevs