What Is GPT-Live-1? OpenAI's Full-Duplex Voice Model Explained

GPTLive1 is an independent site. We are not affiliated with, endorsed by, or connected to OpenAI. This is an explainer, not an official source.

On July 8, 2026, OpenAI shipped GPT-Live — a new generation of voice models designed to make talking to ChatGPT feel like a real conversation instead of a walkie-talkie. Two versions rolled out: GPT-Live-1 and GPT-Live-1 mini.

The one thing that changed: full-duplex

Older voice modes were half-duplex: you talk, it waits, then it talks. GPT-Live-1 is full-duplex — it can listen and speak at the same time. In practice that means it can throw in a quick "mhmm" or "yeah" while you're still talking, jump in for fast back-and-forth, or just stay quiet when you clearly need a second to think.

That sounds small. It isn't. Full-duplex turn-taking is the difference between a stilted exchange and something that feels like a phone call with a person.

It's also smarter

GPT-Live-1 is OpenAI's most capable voice model so far. When a question needs web search, deeper reasoning, or heavier work, it quietly delegates to a frontier model in the background (GPT-5.5 at launch) and folds the answer back into the conversation when it's ready — so you get natural voice and real reasoning.

The benchmark headline: GPT-Live-1 at its highest reasoning setting scored 84.2% on GPQA (a hard scientific-reasoning test), up from 45.3% for the old Advanced Voice Mode. That's not an incremental bump.

Who gets it

• GPT-Live-1 becomes the default voice model for Go, Plus, and Pro users.
• GPT-Live-1 mini becomes the default for Free users.
• Both rolled out to ChatGPT users globally starting launch day.

GPT-Live-1 vs Advanced Voice Mode

	Advanced Voice Mode	GPT-Live-1
Turn-taking	Half-duplex (take turns)	Full-duplex (listen + speak at once)
Interjections	No	Yes ("mhmm", "yeah")
Reasoning (GPQA)	45.3%	84.2%
Background frontier model	No	Yes (GPT-5.5 at launch)

Want to hear natural AI voice right now?

You don't need a ChatGPT subscription to hear what natural AI speech sounds like. Our free in-browser text-to-speech tool lets you type any text and listen to a natural voice in seconds — no install, no account required to try. It's powered by our own voice technology (not GPT-Live), but it's a fast way to get a feel for where conversational voice AI is heading.