OpenAI Brings GPT-Live-1.1 to the API With Lower-Cost Standby
OpenAI's API changelog lists GPT-Live-1.1 for the Live API, with improvements to pronunciation, accent consistency, and multilingual transcription. A new WebRTC standby configuration works with GPT-Live-1 and GPT-Live-1.1: inference pauses during inactivity, then resumes when the user speaks. OpenAI lists standby at $0.02 per minute versus $0.05 for active session time; backend model and tool usage are billed separately.
OpenAI announced that GPT-Live-1.1 is now available in the API. Its API changelog lists the model for v1/live/sessions and says it improves pronunciation, accent consistency, and transcription in multilingual voice conversations. (OpenAI API changelog; GPT-Live-1.1 model page)
The same update adds a standby configuration for GPT-Live-1 and GPT-Live-1.1 WebRTC sessions. Instead of running inference continuously through a quiet period, an application can pause inference after an idle timeout while keeping the session open. (OpenAI session guide)
Configure standby for a WebRTC session
OpenAI's session guide says to set session.wake_policy.type to speech and choose a whole-number session.idle_timeout_s from 5 to 60 seconds. If the timeout is omitted, inference stays active after the first wake until the session ends. (OpenAI session guide)
When the timeout is reached without an input or output transcript, inference pauses. The connection and session ID remain the same, and the conversation history is restored when the user speaks again. The app must keep the microphone track active during standby; the first response after a short utterance may take longer. (OpenAI session guide)
Standby lowers session-time charges
OpenAI lists $0.05 per minute for active GPT-Live session time and $0.02 per minute for standby. The lower standby rate applies to WebRTC sessions while inference is paused. Backend model and tool usage are billed separately, so the standby rate does not represent the full cost of an agent-assisted voice interaction. (GPT-Live-1.1 model page)
This is most relevant to voice apps that need to keep a conversation available through short pauses. Standby keeps the session and its history in place, but developers should account for the possible delay on the first response after a pause and keep their microphone track active. (OpenAI session guide)
OpenAI's ChatGPT release notes list GPT-Live-1.1 and new voices in ChatGPT as an October 15, 2026 update for ChatGPT Voice. This predates the API availability notice in the Nov 19 changelog. For the underlying full-duplex architecture, see our earlier GPT-Live architecture coverage. (OpenAI Help Center release notes; OpenAI API changelog)
Read next
Related coverage- OpenAI
OpenAI Details GPT-Live’s Full-Duplex Voice Architecture
OpenAI says GPT-Live is a third-generation voice system that can listen and speak at the same time, keep media flowing while deeper reasoning and tool use run asynchronously, and power ChatGPT Voice. GPT-Live is now available through the API, where GPT-Live-1.1 supports WebRTC standby during inactivity.
- OpenAI
OpenAI’s GPT-Live-1.1 powers the new ChatGPT Voice experience (web search, memory, and multimodal chat)
OpenAI’s GPT-Live family powers ChatGPT Voice, with GPT-Live-1.1 and GPT-Live-1.1 Quip now rolling out. The October 15 update adds more natural accents, new voice options across six languages plus Bubbie in English, and more varied search-status phrases. Earlier Voice capabilities include full-duplex conversation, web search, memory, and text-and-image chat.
- OpenAI
OpenAI adds GPT-Realtime-2.1-mini to the API with reasoning and tool use
GPT-Realtime-2.1-mini is a distilled reasoning model for realtime voice, with function calling, a 128,000-token context window, and text, audio, and image input. OpenAI's pricing page lists it at the same rates as GPT-Realtime-mini: $0.60/$2.40 per 1M text tokens and $10/$20 per 1M audio tokens.
- OpenAI
OpenAI Lists New ChatGPT Voice in Desktop Pricing for Work and Codex
OpenAI's current Codex pricing documentation lists Voice in Desktop at $0.05 per connected minute and 1.25 credits per minute for credit-based Business, Edu, and Enterprise workspaces. OpenAI Developers amplified a post describing the change as an approximately 60% price reduction for Voice in Work and Codex; the documentation confirms the current rates but does not provide a prior-rate comparison.
Sources
- OpenAI API changelog, November 19, 2026: https://developers.openai.com/api/docs/changelog
- OpenAI, GPT-Live-1.1 model page: https://developers.openai.com/api/docs/models/gpt-live-1.1
- OpenAI, Managing GPT-Live sessions: https://developers.openai.com/api/docs/guides/live-conversations
- OpenAI Help Center, ChatGPT release notes: https://help.openai.com/en/articles/6825453-chatgpt-release-notes
Frequently Asked Questions
What is GPT-Live-1.1?
GPT-Live-1.1 is OpenAI's full-duplex voice model for multilingual conversations. OpenAI says it improves pronunciation, accent consistency, and transcription, and lists it for the Live API at `v1/live/sessions`. ([OpenAI API changelog](https://developers.openai.com/api/docs/changelog); [GPT-Live-1.1 model page](https://developers.openai.com/api/docs/models/gpt-live-1.1))
How does standby work?
For a WebRTC session, set `session.wake_policy.type` to `speech` and `session.idle_timeout_s` to a whole number from 5 to 60. After that period without an input or output transcript, inference pauses while the session connection and history remain available; speech resumes the session. ([OpenAI session guide](https://developers.openai.com/api/docs/guides/live-conversations))
How much does GPT-Live standby cost?
OpenAI lists $0.02 per minute for standby session time and $0.05 per minute for active session time. Backend model and tool usage are billed separately. ([GPT-Live-1.1 model page](https://developers.openai.com/api/docs/models/gpt-live-1.1))
Does standby work with GPT-Live-1 as well?
Yes. OpenAI documents standby for GPT-Live-1 and GPT-Live-1.1 WebRTC sessions. The setting pauses inference during inactivity and resumes when someone speaks. ([OpenAI API changelog](https://developers.openai.com/api/docs/changelog))