GPT-Live 1 now available on AI Gateway
GPT-Live 1 supports simultaneous listening and speaking plus delegation to a separate text model, enabling interruptible voice agents without coupling conversation and deep work.
**GPT-Live 1** is available through Vercel AI Gateway as a full-duplex voice model that can listen while speaking. Its **client delegation** lets an application send deeper work to a separately chosen text model while the voice session continues.
For a voice agent, keep conversational interaction in the live model and explicitly govern delegated tools or text models with permissions, confirmations, and cancellation. Wait for `session-started` before sending audio.
**GPT-Live 1** is available through Vercel AI Gateway as a full-duplex voice model that can listen while speaking. Its **client delegation** lets an application send deeper work to a separately chosen text model while the voice session continues. For a voice agent, keep conversational interaction in the live model and explicitly govern delegated tools or text models with permissions, confirmations, and cancellation. Wait for `session-started` before sending audio. Integration requires **AI SDK 7**, `@ai-sdk/openai` 4.0.67 or later, and a WebSocket client. Delegated requests are billed separately, voice usage continues during them, and the shown snippets omit several production concerns.
This turns GPT-Live-1 into a concrete gateway integration and separates continuous conversation from delegated deep work. Relative to modular transcription pipelines, it offers a managed full-duplex loop, but adds concurrent billing and a new authorization boundary around delegated models and tools. The experimental dependencies and omitted production controls keep it an evaluation path, not a default architecture.