PRODUCT • • 4 min read

OpenAI Puts GPT-Live-1.1 on the API at $0.05 a Minute

By Ultrathink
Thumbnail for: OpenAI Adds GPT-Live-1.1 to the API

OpenAI's API changelog dates GPT-Live-1.1 to November 19, 2026. The model id is gpt-live-1.1. The model page calls it a full-duplex voice model that can listen while it speaks, and says that, compared with GPT-Live-1, it transcribes speech more accurately and holds a more consistent accent and voice across languages. An OpenAI post on the developer forum, timestamped November 20 at 4:12 UTC, says the same model has powered voice in ChatGPT Plus and Pro since October 15.

The numbers

  • Active voice time is $0.05 per minute. Standby is $0.02 per minute. Both figures are on the GPT-Live-1.1 model page and on the API pricing page. The pricing page says GPT-Live bills voice sessions per second.
  • Standby is for WebRTC sessions while inference is paused. The session guide says to set wake policy type to speech, and to set an idle timeout from 5 to 60 seconds. Omit the timeout and inference stays active after the first wake until the session ends. The optional voice-activity threshold runs from 0 to 1 and defaults to 0.5.
  • Creating a WebRTC session bills 15 seconds of voice duration while it initializes. OpenAI's cost guide says that charge is credited once the session is running, and that a 90-second active session already includes those 15 seconds.
  • Backend model and tool use is billed separately. The voice rate does not include it. The printed API rates for GPT-6.2 Astra and Sol are in a separate note.

Where it runs, and where it does not

The model page lists one supported route, v1/live/sessions. Chat Completions, Responses, and the older Realtime API are marked not supported. Supported features on that page are streaming and function calling. Inputs and outputs are audio and text. The Live API reference says the default voice is marin and that the voice cannot be changed after the session starts. The forum post, not the model page, adds that the model is meant to catch precise details such as letters and numbers.

The changelog tells developers to drop gpt-live-1.1 into an existing Live session configuration and to say in plain language when the model should hand a task to the backend. Standby is listed there for both GPT-Live-1 and GPT-Live-1.1 WebRTC sessions. The lower idle rate is not exclusive to 1.1. The pricing table gives both models the same two rates.

What is not disclosed?

The model page says GPT-Live-1.1 and GPT-Live-1 have separate concurrent-session limits, and that each limit applies across inference regions. It does not print the numbers. It points to the limits on the caller's own account. The 1.1 model page does not state a knowledge cutoff. The October 15 date for ChatGPT Plus and Pro is in the November 20 forum post. The changelog entry dated November 19 does not say it. The docs also do not describe a price change from GPT-Live-1. Both rows are $0.05 active and $0.02 on standby.

Quick answers

What happened with OpenAI?
OpenAI's API changelog dates GPT-Live-1.1 to November 19, 2026. The model id is gpt-live-1.1. The model page calls it a full-duplex voice model that can listen while it speaks, and says that, compared with GPT-Live-1, it transcribes speech more accurately and holds a more consistent accent and voice across languages. An OpenAI post on the developer forum, timestamped November 20 at 4:12 UTC, says the same model has powered voice in ChatGPT Plus and Pro since October 15.
What are the key numbers?
Active voice time is $0.05 per minute. Standby is $0.02 per minute. Both figures are on the GPT-Live-1.1 model page and on the API pricing page. The pricing page says GPT-Live bills voice sessions per second.
What is not disclosed?
The model page says GPT-Live-1.1 and GPT-Live-1 have separate concurrent-session limits, and that each limit applies across inference regions. It does not print the numbers. It points to the limits on the caller's own account. The 1.1 model page does not state a knowledge cutoff. The October 15 date for ChatGPT Plus and Pro is in the November 20 forum post. The changelog entry dated November 19 does not say it. The docs also do not describe a price change from GPT-Live-1.

This article was ultrathought.

Pass it on

The story stays open.

The story stays open. Share it once, or start the free brief, to open the sample format.

Sample of the morning brief

Not a live send. The real brief arrives only after you confirm your email.

The example stays hidden until you share or start the brief.

Founding reader means 3 people confirm the brief from your link. The count lives on the subscriber record. This page does not show a live tally.

Related stories

More on OpenAIMore AI Models