OpenAI launched GPT-Live, the new ChatGPT Live voice, on July 8, 2026, and it quietly replaced the Advanced Voice Mode you have been using for the past two years. The headline change is technical but you will feel it in the first ten seconds. GPT-Live is full-duplex, which means it can listen and speak at the same time, so you can cut in mid-sentence and it reacts like a person instead of waiting politely for you to finish.
This guide covers what GPT-Live actually is, the difference between the free GPT-Live-1 mini and the paid GPT-Live-1, how to turn it on, and how it stacks up against Apple’s brand-new Siri AI on your iPhone and Mac. We stick to what OpenAI and independent outlets have confirmed so far, and we flag the parts that are still rough.
The Key Takeaways
- GPT-Live launched July 8, 2026 and replaces Advanced Voice Mode by default for everyone.
- It ships in two versions, GPT-Live-1 mini for free users and the larger GPT-Live-1 for Go, Plus, and Pro subscribers.
- It is full-duplex, so it listens and speaks at once and handles interruptions in real time.
- For hard questions it delegates to GPT-5.5 in the background, then brings the answer back into the conversation.
- It rolled out worldwide on day one with nine remastered voices and live translation.
What Is ChatGPT Live (GPT-Live)?
GPT-Live is OpenAI’s new voice model, launched on July 8, 2026, that can listen and speak at the same time. It replaces the older Advanced Voice Mode inside ChatGPT, interrupts and gets interrupted naturally, and hands off complex questions to a frontier text model in the background. Free users get the smaller GPT-Live-1 mini, while paid plans include the full GPT-Live-1. Because it lives inside the app, many people search for it as ChatGPT Live or the new ChatGPT live voice, and both names point to the same thing.
The word that matters here is full-duplex. Old voice assistants, including the previous ChatGPT voice mode, worked in turns. You spoke, it stopped, then it answered. GPT-Live drops that back-and-forth structure and makes decisions multiple times per second about whether to speak, pause, interrupt, or reach for a tool.
That single change is why early testers describe it as the first voice AI that feels conversational rather than transactional. OpenAI product lead Atty Eleti told reporters he had 30 to 40 minute conversations with it on walks, and framed voice as a future “primary interface to computing.” Whether that holds up long term is an open question, but the demos landed hard.
GPT-Live-1 vs GPT-Live-1 Mini: Is GPT-Live Free?
Yes, GPT-Live is partly free. Every free ChatGPT account now defaults to GPT-Live-1 mini, which quietly took over from Advanced Voice Mode. The larger, smarter GPT-Live-1 is reserved for paid Go, Plus, and Pro subscribers. Both versions are full-duplex and share the core features, so the free tier is actually usable rather than a stripped-down demo.
| Version | Who gets it | Best for |
|---|---|---|
| GPT-Live-1 mini | Free ChatGPT users (default) | Everyday chat, quick questions, hands-free use |
| GPT-Live-1 | Go, Plus, and Pro subscribers | Longer reasoning, smoother delivery, heavier daily use |
GPT-Live-1 mini (free)
The mini model is the one most people will actually meet, because it is the new default on the free plan. It handles casual conversation, dictation-style questions, and hands-free use around the house without a subscription. For search, deeper reasoning, or agentic tasks it still routes the request to a stronger text model, so you are not stuck with a weak brain just because you are on the free tier.
GPT-Live-1 (paid)
The full GPT-Live-1 is the version paid subscribers get on Go, Plus, and Pro. It is the larger model, which in practice means steadier delivery, better handling of long back-and-forth exchanges, and fewer awkward pauses. If you use voice for real work, like talking through code, drafting emails aloud, or long planning sessions, this is the tier that justifies the flow. You can check the current free-plan limits in our guide to whether ChatGPT is free.
What GPT-Live Can Do
The full-duplex architecture is the foundation, but the features built on top are what make GPT-Live feel different in daily use. Here is what stands out at launch.
Natural interruptions and backchanneling
Because it listens while it talks, you can cut GPT-Live off mid-answer and it adjusts instantly instead of talking over you. It also does what humans do in conversation, dropping in small acknowledgements like “mhmm” or “got it” so you know it is following along. Small touches, but they change how the exchange feels.
Background reasoning with GPT-5.5
GPT-Live is a voice model, not a full reasoning engine, and OpenAI is upfront about that split. When you ask something that needs web search, deep reasoning, or agentic work, it sends the query to the latest frontier text model, currently GPT-5.5, and keeps chatting while the answer comes back. You get conversational speed and heavyweight intelligence without choosing between them. This mirrors the model routing we broke down in our ChatGPT 5.6 explainer.
ChatGPT Live translation and nine voices
GPT-Live ships with nine remastered voices and built-in live translation, so you can hold a rough real-time conversation across languages. That makes ChatGPT live translation handy when you travel or take a call in another language, and it is one of the most searched new features. Just treat it as a fast live translator rather than a stand-in for a professional interpreter, since quality is not perfect yet. It also surfaces rich visual cards for things like weather, stocks, and sports while it talks, so voice and screen work together instead of competing.
How to Use ChatGPT Live
There is almost nothing to set up, because GPT-Live is already the default voice experience. If your ChatGPT app is current, you are most likely using it already. Here is the quick path on iPhone, iPad, or Mac.
- Update the ChatGPT app to the latest version from the App Store.
- Open a chat and tap the voice icon in the bottom right.
- Start talking. GPT-Live is active by default, so no toggle is needed.
- Interrupt whenever you want, and let it finish tasks in the background while you keep chatting.
Free accounts land on GPT-Live-1 mini automatically. If you are on Go, Plus, or Pro, you get the larger GPT-Live-1 without changing a single setting.
GPT-Live vs Siri AI
The obvious rival on your iPhone and Mac is Siri AI, the rebuilt assistant Apple previewed for iOS 27. The two solve overlapping problems from opposite directions. GPT-Live is a cloud voice model tuned for conversation, while Siri AI is a system assistant tuned for your device and personal context. The table below lays out the core differences.
| Feature | GPT-Live | Siri AI |
|---|---|---|
| Architecture | Full-duplex, listens and speaks at once | Turn-based, conversational upgrade |
| Free access | Yes, GPT-Live-1 mini | Yes, built into iOS 27 |
| Background model | GPT-5.5 | Apple models plus Gemini |
| Live translation | Yes | Limited |
| Device integration | App-based | Deep, system-wide |
| Personal context | Limited | Messages, email, photos, on-screen |
| Platforms | iPhone, iPad, Mac, web | iPhone 15 Pro and later |
Where GPT-Live wins
GPT-Live is the better pure conversationalist. The full-duplex flow, natural interruptions, and instant delegation to GPT-5.5 make it feel like talking to a sharp assistant rather than issuing commands. It also works the same across iPhone, iPad, Mac, and the web, so your experience does not depend on owning the newest hardware. For brainstorming, tutoring, or thinking out loud, it is ahead.
Where Siri AI wins
Siri AI wins on everything that touches your actual device. It can pull from your messages, email, and photos, understands what is on screen, and triggers system actions in a way an app-based model cannot. It is also free inside iOS 27 with no subscription math. If you want an assistant that runs your phone rather than one that talks well, Siri AI has the edge. For a fuller picture, see our guide on how to use Siri.
The Limitations Worth Knowing
GPT-Live is impressive, but it is two days old and it shows in places. The live translation demo drew the most attention, where the model spoke Hindi with a heavy American accent that sounded unnatural. OpenAI says it is optimized for most spoken languages, without naming which ones, so treat translation as a helpful rough draft rather than a reliable interpreter.
There is also the split-brain design to keep in mind. Because complex questions get routed to GPT-5.5 in the background, the smoothness of a reply can depend on that handoff. Most of the time it is seamless, but it is a moving target while OpenAI tunes the rollout. Expect the specs and behavior here to shift over the coming weeks.
Use GPT-Live and Every Other Model in One App
The catch with GPT-Live is the same catch as always with ChatGPT. The best version sits behind a paid plan, and you are locked into one company’s voice, one company’s reasoning, and one company’s release schedule. If GPT-Live is having an off day, or you want to compare its answer against another model, you are stuck opening more apps and juggling more logins.
Fello AI takes the other approach. It puts leading models from OpenAI, Anthropic, Google, and others in a single Mac and iPhone app, so you can talk to one, generate an image with another, and draft a document with a third without leaving the window. Instead of betting on one assistant, you pick the right model for each task, the same logic behind our roundup of the best ChatGPT alternatives for Mac. You can download Fello AI on the App Store and try the multi-model workflow yourself.
Conclusión
GPT-Live is the biggest change to ChatGPT voice since it launched, and it is already the default on every account. The full-duplex flow makes it the most natural voice AI you can talk to right now, the free GPT-Live-1 mini makes it accessible, and the GPT-5.5 handoff keeps it smart. Against Siri AI it wins on conversation and loses on device integration, so the two can happily live side by side. If you want the flexibility to reach every model, not just one, start with our roundup of the best AI models and go from there.
FAQ
Is GPT-Live free?
Partly. Free ChatGPT users get GPT-Live-1 mini by default, which replaced Advanced Voice Mode. The larger GPT-Live-1 is reserved for paid Go, Plus, and Pro subscribers.
What does full-duplex mean?
Full-duplex means GPT-Live can listen and speak at the same time. It decides many times per second whether to talk, pause, or interrupt, so you can cut in mid-sentence and it reacts naturally.
What model powers GPT-Live?
GPT-Live handles the conversation itself, and for search, reasoning, or agentic tasks it delegates to the latest frontier text model, currently GPT-5.5, then brings the answer back into the chat.
Is GPT-Live better than Siri?
It depends on the job. GPT-Live is the stronger conversationalist thanks to its full-duplex flow, while Siri AI wins on device integration and personal context because it can reach your messages, email, and on-screen content.
How do I turn on GPT-Live?
You do not need to turn it on. Update the ChatGPT app, open a chat, and tap the voice icon. GPT-Live is the default voice experience, so free users get the mini model and paid users get the full model automatically.
