Models ·
OpenAI Releases GPT-Live-1.1 for Multilingual Voice
Models ·
OpenAI Releases GPT-Live-1.1 for Multilingual Voice
AI summary · Generated from this article
OpenAI has introduced GPT-Live-1.1, a full-duplex voice model designed to improve pronunciation, accent consistency, and transcription accuracy across multilingual conversations. Developers can implement this by selecting the new identifier within existing Live session configurations, though it remains incompatible with the Realtime API. Alongside this, OpenAI launched a standby mode for WebRTC sessions that reduces costs by 60 percent during inactivity. Because the model upgrade and standby feature affect different operational metrics, teams should test them independently to accurately measure improvements in conversational quality and potential cost savings without introducing unforeseen regressions.