OpenAI added GPT-Live-1.1 (gpt-live-1.1) to the Live API on November 19. Developers can select it in an existing Live session configuration. OpenAI says the model improves pronunciation, accent consistency, and transcription in multilingual conversations. Voice sessions cost $0.05 per active minute. WebRTC sessions cost $0.02 per minute in standby while inference is paused. Backend model and tool usage is billed separately.
GPT-Live-1.1 can listen while speaking, and developers can describe in plain language when it should hand work to their backend for reasoning or tools. The newly added standby setting also works with GPT-Live-1, pausing inference during inactivity and resuming when someone speaks. That gives voice applications an explicit way to manage quiet periods during an ongoing connection.
