GPT-Live 1.1 now available on AI Gateway
GPT-Live 1.1 adds more accurate transcription and steadier voice across languages, with the same Live endpoint and delegation flow. Existing integrations can switch by changing the model ID.
**GPT-Live 1.1** supports full-duplex voice: callers can interrupt or add detail while it speaks. Vercel says transcription is more accurate and voice and accent stay more consistent across languages than in the previous model.
Existing integrations use the same Live endpoint and client delegation; switch to the new model ID. Delegated work can run through another AI Gateway text model while the voice session continues.
**GPT-Live 1.1** supports full-duplex voice: callers can interrupt or add detail while it speaks. Vercel says transcription is more accurate and voice and accent stay more consistent across languages than in the previous model. Existing integrations use the same Live endpoint and client delegation; switch to the new model ID. Delegated work can run through another AI Gateway text model while the voice session continues. Your app still handles delegated work, permissions, confirmations, and cancellation. Delegated requests are billed separately, and the examples omit connection, audio streaming, and shutdown details.
GPT-Live 1.1 keeps the existing Live endpoint and delegation pattern while claiming better transcription accuracy and more consistent voice and accent across languages. Delegation still needs app-level permissions, confirmations, and cancellation, and incurs separate billing; the announcement omits connection, streaming, and shutdown details.