GPT-Live-1.1 arrives in the API: in standby, a WebRTC pause costs $0.02 a minute
OpenAI has opened GPT-Live-1.1 to developers. The voice model has powered voice in ChatGPT Plus and Pro since 15 October. In the new standby mode, a silent WebRTC call costs $0.02 a minute instead of the usual $0.05.
GPT-Live-1.1 listens even while it speaks. OpenAI announced its arrival in the API on its X account on 19 November and on the model’s documentation page.
OpenAI said the new model transcribes speech more accurately than GPT-Live-1 and keeps a more consistent accent and voice across languages. The model reliably understands letters and numbers and follows instructions, the company said. It also completes tasks more reliably than before. These improvements are OpenAI’s own claims: the company gives no figures for them, and the model page lists performance and speed as unspecified.
Voice sessions are billed by duration. An active minute costs $0.05. In standby, the developer sets an idle timeout of 5–60 seconds. Once a call has been silent for that long, the model pauses inference. The WebRTC call nevertheless stays open, at $0.02 a minute, until someone speaks again. A ten-minute pause in a WebRTC call, for example, costs $0.20 rather than $0.50.
Developers can also enable standby in GPT-Live-1.
The per-minute price does not cover the full cost of a voice agent. The developer describes in plain language when the model should hand a task to a backend system for reasoning or tool use. Developers can choose the backend model themselves; use of that model and of tools is billed separately.
The model accepts and produces audio and text, but not images or video. It supports function calling but not structured outputs or fine-tuning.
The model ID is gpt-live-1.1. GPT-Live-1.1 and GPT-Live-1 have separate concurrent-session limits, and each limit is shared across all inference regions.