Free guide
GPT-Live-1.1 Standby: Stop Paying Full Price for Silence
OpenAI's voice model can now pause while nobody is talking and bill a lower rate until someone speaks. What changed, what it costs, how to turn it on, and the fine print.

The short version
- GPT-Live-1.1 came to OpenAI's API on Nov 19, 2026, with more accurate transcription and a more consistent accent and voice across languages.
- The same day OpenAI added standby for GPT-Live-1 and GPT-Live-1.1 WebRTC sessions: $0.02 a minute while paused vs $0.05 a minute active.
- Without an idle timeout, inference stays active until the session ends, and silence bills at the active rate.
How it actually works
You pay for time, not tokens
GPT-Live voice sessions are billed by duration, per second, with no rounding up to the next minute. Backend models and tools you delegate to are billed separately. OpenAI's docs say active time includes speech, silence before the idle timeout, and time waiting on backend work while inference is running.
The interpreter who naps
Think of an interpreter on a per-minute rate. Before standby, a quiet room still cost the full rate. With standby, after the quiet stretch you pick, the model pauses, the line stays open and the meter drops from 5 cents to 2 cents a minute. When someone speaks, it wakes up and picks the conversation back up.
What stays on while it naps
Keep the microphone track active. The connection and session ID stay the same, and the session restores its conversation history when the user speaks. Your app gets a session.paused event when it goes to standby and session.resumed when inference is ready again.
Muting is not standby
Muting microphone input leaves the session active, so it still bills the active rate. For long gaps where your app can disconnect, OpenAI suggests closing the session and starting a new one when the user returns.
How to set it up
1.Start from OpenAI's WebRTC quickstart
Standby only works on WebRTC sessions created with POST /v1/live/sessions. OpenAI's Voice WebRTC guide has a full Node and Python example that creates a session with model gpt-live-1.1. Get that running first.
https://developers.openai.com/api/docs/guides/voice-webrtc2.Add the two standby fields to the session
In the session object you send when you create the WebRTC session, set wake_policy.type to speech and idle_timeout_s to a whole number from 5 to 60. If you leave the timeout out, inference stays active after the first wake until the session ends.
session: { model: "gpt-live-1.1", wake_policy: { type: "speech" }, idle_timeout_s: 20 }3.Optional: tune how easily it wakes
wake_policy.vad_threshold takes a number from 0 to 1 (default 0.5). Higher values need stronger evidence of speech before it wakes, which helps in noisy rooms.
wake_policy: { type: "speech", vad_threshold: 0.6 }4.Listen for paused and resumed
Watch for session.paused and session.resumed on the sideband, or on the WebRTC data channel if its event settings allow them. Show a small listening state in your UI while it's paused, and keep the mic track live.
The fine print
- Prices from OpenAI's model page: $0.05 a minute active, $0.02 a minute standby, for GPT-Live-1 and GPT-Live-1.1. Backend model and tool usage is billed separately.
- Standby is for WebRTC sessions only.
- After standby, the first response to a short utterance may take longer, per OpenAI's docs.
- Creating a WebRTC session bills 15 seconds of voice time while it initializes (credited once it runs), so closing and reopening sessions often has its own cost.
- GPT-Live-1.1 and GPT-Live-1 have separate concurrent-session limits. Check your organization's limits page.
FAQ
What is GPT-Live-1.1?
It's OpenAI's full-duplex voice model for the API, meaning it can listen while it speaks. Compared with GPT-Live-1 it transcribes speech more accurately and keeps a more consistent accent and voice across languages. The model ID is gpt-live-1.1 on the v1/live/sessions endpoint.
How much does GPT-Live-1.1 cost?
Voice sessions are billed by duration: $0.05 a minute while active and $0.02 a minute on standby. Charges accrue per second. Any backend model or tool calls are billed separately at their normal rates.
Does standby work on GPT-Live-1 too?
Yes. OpenAI's Nov 19 changelog says standby works for GPT-Live-1 and GPT-Live-1.1 WebRTC sessions.
Does the user have to press a button to wake it?
No. With wake_policy set to speech, the server listens with voice activity detection and resumes inference when someone speaks. The conversation history comes back with it.
Is there a downside to standby?
The first reply to a short utterance after a pause may take a little longer. And silence before the idle timeout still bills at the active rate, so a 60 second timeout saves less than a 10 second one.
Sources
- GPT-Live-1.1 model page (pricing)
- Configure standby
- GPT-Live usage and costs
- OpenAI API changelog
- @OpenAIDevs launch post
Facts checked 2026-11-20 against OpenAI's GPT-Live-1.1 model page, the Live conversations guide (Configure standby), the GPT-Live usage and costs guide, the API changelog (Nov 19) and the @OpenAIDevs launch post.
Want help building with tools like this?
Join the Shipping Skool community free. Paid members get live bootcamp calls every week and weekly-updated courses.
Join Shipping SkoolMore free guides
Chrome DevTools MCP: Let Your Coding Agent See the Page It Builds
Google's free, open-source way to give your coding agent a real Chrome window and the DevTools you already use. What it does, how it works, how to set it up, and the fine print.
GPT-6.2 Drops Its Reasoning When You Edit Old Messages: How to Keep It
Every reasoning item is now tied to the conversation before it. Edit that history and the API drops the reasoning that follows. Here's how it works and how to keep it.
Inkling 2 Mini: Run Thinking Machines' Looped AI Model on Your Mac
Thinking Machines' new open-weights model runs on a 32GB Mac with the wifi off. What it is, how the loop trick works, how to set it up, and the fine print.