OpenAI on November 19, 2026 added GPT-Live-1.1 to v1/live/sessions and introduced standby billing for GPT-Live WebRTC calls, according to its API changelog.
The standby mode gives voice apps a lower-cost state when a call goes quiet. Instead of ending the connection, a session can stay open on WebRTC, pause inference during inactivity, and resume when someone speaks. OpenAI says the option works with both GPT-Live-1 and GPT-Live-1.1.
Pricing on the model page lists $0.05 per active minute for GPT-Live-1.1 voice sessions and $0.02 per standby minute.
OpenAI also frames GPT-Live-1.1 as an audio-quality update, citing better pronunciation, more consistent accents, and stronger transcription in multilingual voice conversations. The model is supported on v1/live/sessions, not the Responses, Chat Completions, or Realtime endpoints.
For teams building phone or voice agents, the practical change is cost control during silence: idle-timeout and session-lifecycle choices can now directly affect call spend.
