OpenAI on November 19, 2026 added Standby to the GPT-Live API and said GPT-Live-1.1 is available in the API. Standby pauses inference during inactivity on WebRTC sessions and resumes it when someone speaks. Official pricing for both gpt-live-1 and gpt-live-1.1 lists Active time at $0.05 per minute and Standby at $0.02 per minute.
Key Facts
| Item | Detail |
|---|---|
| Date | November 19, 2026, in the OpenAI API Changelog (Feature, model gpt-live-1.1, API v1/live/sessions). The verified OpenAI Developers account, @OpenAIDevs, posted about the release at 20:00:32 UTC the same day. |
| GPT-Live-1.1 | The changelog said GPT-Live-1.1 improves pronunciation, accent consistency, and transcription in multilingual voice conversations. Developers can use gpt-live-1.1 in an existing Live session configuration. The changelog also said to describe in plain language when to delegate to a backend. |
| Standby | A new configuration for GPT-Live-1 and GPT-Live-1.1 WebRTC sessions. Official docs say to use Standby to keep a WebRTC connection open during inactivity and resume inference when someone speaks. For longer gaps, when the application can disconnect, the docs say to close the session and start another when the user returns. |
| Pricing | For gpt-live-1.1 and gpt-live-1, Active is $0.05 per minute and Standby is $0.02 per minute. Official pricing pages say the lower standby rate applies to WebRTC sessions while inference is paused. The gpt-live-1 model page says voice sessions are billed per second, and that session duration is not rounded up to the next whole minute. Backend model and tool usage is billed separately. |
| How to configure it | Set session.wake_policy.type to speech so the session waits for speech before starting inference. session.idle_timeout_s is an integer from 5 to 60. After that many seconds without an input or output transcript, the session returns to standby. If the timeout is omitted, inference stays active after the first wake until the session ends. An optional session.wake_policy.vad_threshold ranges from 0 to 1 and defaults to 0.5. The threshold applies on the first wake and after each pause. |
| While in standby | The microphone track must stay active. The connection and session ID do not change. Conversation history is restored on resume. After standby, the first reply to a short utterance may be slower. The docs say GPT-Live bills standby time at a lower rate than active time. |
| Events | session.resumed means inference is ready, including on the first wake. session.paused means an idle timeout has stopped inference and the session has entered standby. The connection and sideband remain open. |
Background
GPT-Live is OpenAI's live voice API, reached at v1/live/sessions. GPT-Live-1 was already listed in the API. The November 19 changelog presented GPT-Live-1.1 and Standby together.
A post in the OpenAI developer Community Announcements, written by @sps and created November 20, 2026 at 4:12 a.m. UTC, said GPT-Live-1.1 has powered voice in ChatGPT Plus and Pro since October 15. That sentence appears in the Community announcement, not in the API changelog. This article records it with that source, and does not treat it as changelog text.
@OpenAIDevs wrote: "GPT-Live-1.1 is now available in the API. It speaks with more natural and consistent accents across languages, supports lower idle costs with a new standby option, and completes tasks even more reliably." A follow-up post at 20:00:39 UTC began, "Standby is a new configuration in the GPT-Live API that lets developers manage periods of inactivity without interrupting the conversation. It's available for both GPT-Live-1 and GPT-Live-1.1. Set an idle timeout of 5-60 seconds. After that period of silence, the model" The rest of that post was truncated in the text available for this article. The configuration and billing details above follow the official documentation and the Community announcement.
What We Don't Know Yet
- Official documentation describes Standby as a configuration for WebRTC sessions.
- The second @OpenAIDevs post was truncated in the available text. This article does not complete that sentence.
- The sources reviewed for this article did not include a broader GPT-Live product roadmap beyond this changelog.
Sources
- OpenAI API Changelog, November 19, 2026
- Managing GPT-Live sessions, including the sections on managing idle sessions and configuring Standby
- API pricing
- gpt-live-1.1 model page
- gpt-live-1 model page
- Live API reference
- GPT-Live-1.1 has arrived in the API, OpenAI developer Community announcement by @sps, created November 20, 2026 at 4:12 a.m. UTC
- @OpenAIDevs, November 19, 2026, 20:00:32 UTC
- @OpenAIDevs follow-up, November 19, 2026, 20:00:39 UTC
- Digg summary, used to cross-check official figures