Vercel AI Gateway adds GPT-Live 1.1
Vercel AI Gateway now supports OpenAI's full-duplex GPT-Live 1.1 voice model, enabling real-time conversational streaming with per-second billing and no provider markup. Developers can integrate live audio via the AI SDK with WebSocket support and client-side model delegation for background tasks.
Bringing full-duplex voice to unified inference gateways lowers the barrier for production voice agents from painful WebRTC infrastructure to simple API calls. The client delegation pattern here is especially important, establishing a standard architecture where fast voice models handle speech while specialized LLMs handle reasoning in the background.
- –Zero-markup per-second billing avoids minimum session commitments, making voice experimentation viable for smaller teams.
- –Native client delegation allows GPT-Live 1.1 to offload complex tool calls or reasoning to text models like GPT-6.2-sol over a commentary channel without breaking speech continuity.
- –Model-level upgrades in GPT-Live 1.1 reduce voice timbre drift and improve transcription accuracy across non-English languages.
- –WebSocket orchestration through the Vercel AI SDK eliminates the need to manage standalone provider credentials or custom streaming middleware.
DISCOVERED
21m ago
2026-11-20
PUBLISHED
55m ago
2026-11-20
RELEVANCE
AUTHOR
vercel_dev