The development
What happened
OpenAI released GPT-Live-1 in its API for voice applications. The model can listen while speaking and keep a conversation moving while a separate backend handles reasoning and tools.
Key facts
OpenAI announced the API release on 10 September 2026.
Full-duplex operation allows simultaneous listening and speaking, including responses to interruptions.
Developers choose a backend model or agent for delegated reasoning and tool use.
The release includes telephony support and prompt control of voice style; custom voice eligibility requires contacting sales.
The context
Why it matters
The architecture separates voice conversation from backend task execution, allowing interaction to continue during tool work. Developers still need to test their own application, permissions and task outcomes; publisher examples do not prove reliability for every deployment.