OpenAI's new full-duplex voice models for ChatGPT — they listen and speak at the same time, and can hand a question off for a web search mid-conversation.
Last verified · 2026-07-08 · by Moe Ameen
GPT-Live is a new generation of voice models OpenAI began rolling out on July 8, 2026 to power ChatGPT Voice. The headline change is that they are full-duplex: the model can take in audio and produce speech at the same time, so it can drop in a quick "mhmm" or "got it," trade fast back-and-forth, interrupt less, and stay quiet when you pause to think. The aim is a conversation that feels closer to talking to a person than to dictating commands and waiting for a reply.
Two versions shipped together. GPT-Live-1 becomes the default model behind ChatGPT Voice for Go, Plus, and Pro subscribers; a smaller GPT-Live-1 mini becomes the default for free users. Both reach ChatGPT users worldwide across iOS, Android, and the web. OpenAI has said more than 150 million people already talk to ChatGPT every week through its Voice and Dictation features, which is the install base this upgrade lands on.
The other notable piece is that voice is no longer isolated from search and reasoning. Mid-conversation, GPT-Live can hand a question to a frontier model such as GPT-5.5 for a web search or heavier reasoning, then keep talking. You can set the reasoning level to fit the moment — Instant for fast replies, or Medium and High when you want it to think longer — and ChatGPT Voice can surface visual cards for things like weather, stocks, and sports while you talk. It continues to support search, memory, images, and file uploads.
Honest framing: GPT-Live is a conversational interface, not a content-production tool. It is excellent for talking through ideas, asking questions hands-free, and getting spoken answers backed by live search — but it outputs a conversation, not on-brand posts, captioned video, carousels, or a publishing schedule. Video and screen sharing are not in this release; OpenAI says both are being worked on. There is no separate GPT-Live price — access comes with your ChatGPT plan (Free, Go, Plus, or Pro).
GPT-Live is the microphone in front of the machine — it is where the idea gets said out loud, not where it becomes a post. Its real strength is capture and interrogation: you talk through a hook while you're walking, it pushes back, you ask it to look something up, it web-searches through GPT-5.5 and reads the answer back. What you walk away with is a spoken transcript and a clearer take — raw material, not finished content. Kompozy is the layer that turns that spoken idea into a week of published work. Paste the transcript or the sharpened angle into Kompozy's Quick Ingest and one voice note fans out into a Blog Article, a Carousel, native Text Posts, and an Email Newsletter, all held to your voice by the Persona Brief and banned-word filters so a rambling brainstorm reads as your brand, not a raw dictation.
The point is that talking is fast but distribution is the work GPT-Live doesn't do. It writes no caption, keeps no brand voice across formats, builds no video, and posts to nothing. Kompozy generates the formats a voice model can't stage at all — Persona Shorts and HeyGen avatar video that put a face-locked recurring identity on camera reading the script you just dictated, Clipped Shorts, brand-exact HyperFrames carousels — and then does the part no assistant touches: schedules and publishes the whole set across nine social platforms plus blog and email from one queue, with Autopilot and a per-post review pipeline. Say the idea into GPT-Live; ship it everywhere with Kompozy.
GPT-Live is a new generation of OpenAI voice models that power ChatGPT Voice, rolled out starting July 8, 2026. They are full-duplex — they can listen and speak at the same time — so conversations feel more natural, with quick back-and-forth, fewer interruptions, and pauses when you need to think.
The main difference is the full-duplex architecture: GPT-Live can take in audio and produce speech simultaneously rather than strictly taking turns. It can also hand a question mid-conversation to a frontier model like GPT-5.5 for a web search or deeper reasoning, and it lets you set an Instant, Medium, or High reasoning level.
GPT-Live-1 is the default model behind ChatGPT Voice for Go, Plus, and Pro users; GPT-Live-1 mini is the default for free users. It rolled out to ChatGPT users worldwide on iOS, Android, and the web. There is no separate GPT-Live price — access comes with your ChatGPT plan.
No. GPT-Live is a conversational voice interface — it talks, answers, and can search the web, but it does not produce captioned video, carousels, blogs, or a publishing schedule, and video and screen sharing are not in this release. To turn a spoken idea into finished, on-brand posts across platforms, capture the transcript and run it through a content engine like Kompozy.
Yes. During a conversation it can hand a question to a frontier model such as GPT-5.5 for a web search or heavier reasoning, then keep talking. ChatGPT Voice can also show visual cards for things like weather, stocks, and sports while you speak.