Breaking
OpenAI launches GPT-Live-1 for API voice agents
OpenAI released GPT-Live-1 through its API on September 10. The voice model can listen and speak at the same time. It can handle interruptions and send deeper reasoning or tool calls to a separate text model. In an early evaluation, Speak reported almost 80% fewer interruptions than with its previous turn-based system. OpenAI cited the customer evaluation in its launch post.
What happened
OpenAI launched GPT-Live-1 for developers building voice apps and phone agents. OpenAI says one model handles incoming and outgoing audio at the same time. This differs from a chain of speech-to-text, a text model, and text-to-speech systems. GPT-Live-1 can send reasoning and tool calls to another model. Prompts can shape its tone, pace, and speaking style.
Why it matters now
OpenAI is making full-duplex speech available in third-party products through its API. Teams can pair fast conversation with a separate reasoning model, giving them another way to build voice agents. The interruption result comes from Speak's early evaluation. Wider deployments still need to test latency, interruption quality, and reliability during long sessions.
Updates
What changed
OpenAI launches GPT-Live-1 for API voice agents
OpenAI released GPT-Live-1 through its API. The full-duplex voice model can handle interruptions while speaking. It can also send deeper reasoning or tool calls to another model. In an early evaluation, Speak reported almost 80% fewer interruptions than with its previous turn-based system.
Verification
Primary evidence before publication.
Social chatter can identify a lead. It does not authorize a HotTea live story.