GPT-Live-1 Explained: OpenAI's New Full-Duplex Voice Model
OpenAI launched GPT-Live on July 8, 2026, a new generation of voice models that make talking to ChatGPT feel far more like a real conversation. The headline feature is full duplex, meaning the model can listen and speak at the same time instead of waiting for you to finish. Here is what GPT-Live-1 does, how it compares to the old Advanced Voice Mode, and who gets access.
What is GPT-Live-1?
GPT-Live-1 is OpenAI's newest voice model, now powering the ChatGPT Voice experience. It uses a full duplex architecture that processes what you say while it is speaking, so it can jump in with quick acknowledgements like "mhmm", let you interrupt naturally, or stay quiet while you think. It ships in two versions: GPT-Live-1 for paid users and GPT-Live-1 mini for free users.
Full duplex: the core change
Older voice modes were turn based. They waited for a pause, then processed your whole turn, which meant awkward interruptions when background noise or a short pause was mistaken for you finishing. GPT-Live makes interaction decisions many times per second, choosing whether to speak, listen, pause, interrupt, or use a tool. That continuous processing is what enables smoother back and forth and features like live translation.
It delegates the hard work
GPT-Live keeps the conversation flowing while handing off anything that needs web search, deeper reasoning, or agent style steps to a stronger model in the background, which is GPT-5.5 at launch. The result comes back into the live chat when it is ready, so you are not left waiting in silence. OpenAI says it will keep swapping in newer frontier models behind GPT-Live over time.
GPT-Live-1 vs Advanced Voice Mode
In OpenAI's own testing, the improvement is large. At its highest reasoning setting, GPT-Live-1 scored about 84 percent on the GPQA science reasoning test, up from around 45 percent for Advanced Voice Mode, and it scored far higher on an agent web search benchmark. In head to head human tests, people preferred GPT-Live-1 over the old mode most of the time. It also adds visual cards for things like weather, stocks, and sports, and offers three reasoning levels: Instant, Medium, and High.
Who can use it, and what is missing
GPT-Live-1 is the default voice model for Go, Plus, and Pro users, while GPT-Live-1 mini is the default for free users, replacing Advanced Voice Mode. It is rolling out globally on iOS, Android, and the web. At launch it does not yet support video, screen sharing, connected apps, or custom GPTs, and API access is coming later. It uses a fixed set of predefined voices with safeguards against imitating real people.
Where this is heading
OpenAI is positioning voice as a future interface for longer, more complex AI work, not just quick questions. As voice becomes a front end for agents that search, reason, and complete tasks, the model behind the conversation matters more than ever. Keeping an eye on which frontier model powers your tools, and being able to compare them, is worth doing.
Chatgbot lets you work with the latest text models like GPT and Claude in one app, so you can compare the reasoning that increasingly sits behind these new experiences.
Compare the latest AI models in one app
FAQ
What is GPT-Live-1? GPT-Live-1 is OpenAI's new full duplex voice model that can listen and speak at the same time, now powering ChatGPT Voice for paid users. A mini version serves free users.
How is GPT-Live different from Advanced Voice Mode? Advanced Voice Mode was turn based and waited for pauses. GPT-Live processes audio continuously, allowing natural interruptions, backchannel cues, and live translation, and it scores much higher on OpenAI's benchmarks.
Which model powers GPT-Live? GPT-Live handles the live conversation and delegates search and deeper reasoning to a frontier model in the background, which is GPT-5.5 at launch.
Who can use GPT-Live-1? It is the default voice model for Go, Plus, and Pro users, with GPT-Live-1 mini for free users, rolling out globally on iOS, Android, and the web.
What can't GPT-Live do yet? At launch it does not support video, screen sharing, connected apps, or custom GPTs, and API access is planned for later.