OpenAI has begun rolling out GPT-Live, a new full-duplex voice system for ChatGPT that can listen and speak at the same time while handing web searches and harder reasoning tasks to a separate frontier model in the background.
The update started July 8, according to OpenAI, and is arriving for consumer ChatGPT users on ChatGPT.com and the iOS and Android apps. GPT-Live-1 is the default for paid Go, Plus, and Pro accounts; Free accounts receive GPT-Live-1 mini. Availability can still vary by region during the rollout.
For Windows users, the immediate change is in ChatGPT’s web experience rather than the native desktop app. OpenAI says GPT-Live is not initially available in the ChatGPT desktop app, Temporary Chats, custom GPTs, Work, Codex, or connected-app and plugin scenarios.
Previous ChatGPT voice modes generally worked in turns: the user spoke, the system decided the turn had ended, then it replied. GPT-Live instead runs continuously, allowing interruptions, short pauses, acknowledgements, and faster back-and-forth without treating every silence as the end of a request.
More significantly, it can route a question requiring current information, web search, or deeper reasoning to OpenAI’s frontier models. At launch, OpenAI says that background work is handled by GPT-5.5, while GPT-Live maintains the spoken conversation and returns the result when ready.
That moves ChatGPT Search further into hands-free use. A user can ask for time-sensitive information aloud rather than switch to a browser search box, and the conversation can remain active while the query is processed. OpenAI also says the interface can show visual widgets for information such as weather, stocks, and sports alongside spoken responses.
The company says spoken responses appear in the associated ChatGPT chat as streamed text, which should make it easier to review an answer or inspect search results after the conversation.
Those are product claims, not a guarantee of perfect transcription or conversational behavior. OpenAI’s Voice documentation still warns that conversations can be inaccurate, especially where speech overlaps, noise is present, or the exchange moves quickly. Important time-, location-, or date-sensitive answers should still be checked, and Voice uses the device or browser time zone to interpret terms such as “today.”
OpenAI also says it has added audio-specific safety testing and controls for sensitive conversations. That work is relevant to the expanded use case, but it does not change the standard advice for organizations: do not treat a conversational AI response as an authoritative source for operational, legal, medical, or security decisions.
On Windows, the practical route is currently ChatGPT in a supported browser. Consumer users who depend on screen sharing, the desktop app, or custom GPT workflows should not assume the new voice mode replaces their existing setup.
The update started July 8, according to OpenAI, and is arriving for consumer ChatGPT users on ChatGPT.com and the iOS and Android apps. GPT-Live-1 is the default for paid Go, Plus, and Pro accounts; Free accounts receive GPT-Live-1 mini. Availability can still vary by region during the rollout.
For Windows users, the immediate change is in ChatGPT’s web experience rather than the native desktop app. OpenAI says GPT-Live is not initially available in the ChatGPT desktop app, Temporary Chats, custom GPTs, Work, Codex, or connected-app and plugin scenarios.
Voice can now search while the conversation continues
Previous ChatGPT voice modes generally worked in turns: the user spoke, the system decided the turn had ended, then it replied. GPT-Live instead runs continuously, allowing interruptions, short pauses, acknowledgements, and faster back-and-forth without treating every silence as the end of a request.More significantly, it can route a question requiring current information, web search, or deeper reasoning to OpenAI’s frontier models. At launch, OpenAI says that background work is handled by GPT-5.5, while GPT-Live maintains the spoken conversation and returns the result when ready.
That moves ChatGPT Search further into hands-free use. A user can ask for time-sensitive information aloud rather than switch to a browser search box, and the conversation can remain active while the query is processed. OpenAI also says the interface can show visual widgets for information such as weather, stocks, and sports alongside spoken responses.
The company says spoken responses appear in the associated ChatGPT chat as streamed text, which should make it easier to review an answer or inspect search results after the conversation.
Better turn-taking, with familiar limits
The full-duplex architecture is intended to reduce a familiar voice-assistant failure mode: a system cutting in during a brief pause, or missing an interruption because it has already started responding. OpenAI says GPT-Live can wait when asked to listen, respond to interruptions, and better distinguish a user’s speech from background noise.Those are product claims, not a guarantee of perfect transcription or conversational behavior. OpenAI’s Voice documentation still warns that conversations can be inaccurate, especially where speech overlaps, noise is present, or the exchange moves quickly. Important time-, location-, or date-sensitive answers should still be checked, and Voice uses the device or browser time zone to interpret terms such as “today.”
OpenAI also says it has added audio-specific safety testing and controls for sensitive conversations. That work is relevant to the expanded use case, but it does not change the standard advice for organizations: do not treat a conversational AI response as an authoritative source for operational, legal, medical, or security decisions.
What admins and users should know
GPT-Live is not available at launch in ChatGPT Business, Enterprise, or Edu workspaces. Those users retain the existing voice options where available. The new Live mode also does not yet support video or screen sharing; users who need those functions can select Advanced Voice Mode instead.On Windows, the practical route is currently ChatGPT in a supported browser. Consumer users who depend on screen sharing, the desktop app, or custom GPT workflows should not assume the new voice mode replaces their existing setup.
References
- Primary source: ALM Corp
Published: Mon, 13 Jul 2026 13:35:52 GMT
- Independent coverage: thelec.net
Published: 2026-07-13T02:00:46+00:00
OpenAI Upgrades ChatGPT Voice With GPT-Live - The Elec Inc.
OpenAI has upgraded ChatGPT Voice to run on its next-generation voice model, GPT-Live. Thewww.thelec.net - Official source: openai.com
Introducing GPT-Live | OpenAI
A new generation of voice models for natural human-AI interaction, now powering ChatGPT Voice.openai.com - Official source: help.openai.com
ChatGPT — Release Notes | OpenAI Help Center
A changelog of the latest updates and release notes for ChatGPT
help.openai.com
- Official source: deploymentsafety.openai.com
