Voice AI & Customer Operations
OpenAI releases GPT-Live-1 as a full-duplex voice layer
The API model can listen and speak at the same time, delegate deeper work and support telephony; the advertised front-end voice price excludes its backend model and tools.
What happened
OpenAI released GPT-Live-1 in the API as a full-duplex voice model that can process incoming and outgoing audio together, delegate reasoning and tool calls to a backend model, and support telephone-based agents. OpenAI lists the front-end voice layer at $0.05 per minute and says custom voice access requires a separate sales process.
Why it matters for AI
A simultaneous listening-and-speaking layer can make interruption handling more natural and simplify a voice stack. The complete service still needs telephony transport, a backend model, tools, consent and recording rules, escalation, monitoring and a reliable human fallback.
AIRunsThis View
Our interpretation: price a complete call, not just the voice layer. A useful pilot should measure connected-call minutes, backend tokens, telephony fees, tool actions, false interruptions, handoff success and failure recovery on scripted customer scenarios.
What this could mean for us
Relevant to the proposed customer assistant and telephone route, but only as a component. No phone number, account, custom voice, paid API usage or customer call was activated.
Source and method
This item is based on a primary source. AIRunsThis separates the source’s factual announcement from its own interpretation. Vendor performance claims have not been independently reproduced unless stated.