OpenAI's GPT-Live-1.1 model employs a time-based pricing structure rather than token-based. It charges $0.05 per active minute, during which inference is running, and $0.02 per minute for standby time, where the session is connected but paused. Backend model and tool calls are billed separately. The system bills per second and does not round up to the next minute, with a 15-second initial charge for WebRTC sessions that is credited once the session is active. Standby mode requires specific configuration, including a "speech" wake policy and an idle timeout, and is only available for WebRTC sessions. AI
IMPACT Understanding GPT-Live-1.1's time-based pricing is crucial for developers to manage costs effectively, especially by optimizing active vs. standby minutes and backend calls.
RANK_REASON This article details the pricing structure and usage of a specific AI product, GPT-Live-1.1, focusing on its billing mechanics for voice sessions.
AI-generated summary · Google Gemini · from 6 sources. How we write summaries →