Instant
Prioritizes fast conversational replies for everyday questions and fluid back-and-forth.
GPT-Live explained
What is GPT-Live?
GPT-Live is OpenAI's newest ChatGPT Voice experience. It is designed for natural full-duplex conversation: you can speak while it responds, mix voice with text or images, and let harder questions use a higher intelligence level.
Last verified:
GPT-Live is a ChatGPT product experience
GPT-Live currently appears as the Live option under Settings → Voice. Paid plans use GPT-Live-1, while Free uses GPT-Live-1 mini. OpenAI's launch material says GPT-Live was built for fast, natural dialogue and can delegate harder work to a stronger background model.
GPT-Live-1 is not generally available as an API model. The official API notification form still asks developers to sign up to be notified. Read the current GPT-Live API status before choosing an implementation path.
GPT-Live-1 vs GPT-Live-1 mini
The two names describe the ChatGPT Voice models available at launch, not public API model IDs.
| Model | ChatGPT access | Best fit | Current API status |
|---|---|---|---|
| GPT-Live-1 | Paid plans | Higher-quality Live conversation with selectable intelligence where available | Notification form only |
| GPT-Live-1 mini | Free; paid-plan fallback or separate allowance | Faster, lighter Live use and additional allowance | No published GPT-Live API model ID |
Instant, Medium, and High intelligence
If your account exposes Voice intelligence controls, choose under Settings → Voice → Intelligence.
Medium
Adds more reasoning for harder questions, with a possible response-time tradeoff.
High
Uses the strongest available handling for difficult tasks and may take longer, especially with web search.
Nine selectable voices
The current Voice settings list Arbor, Breeze, Cove, Ember, Juniper, Maple, Sol, Spruce, and Vale. Changing voice during a conversation starts a new voice call in the same chat. You can separately set your preferred spoken language.
Live works with text and supported images in the same chat, can continue in the background, and supports CarPlay on eligible iPhones. It does not support video or screen sharing at launch; switch to Advanced Voice for those eligible mobile features.
Important launch limits
Availability still varies by plan, region, workspace, and app version. Since July 23, 2026, GPT-Live-powered ChatGPT Voice is available in Work and Codex in the macOS and Windows desktop apps; standalone Work/Codex Voice is not offered on web or mobile.
Use GPT-Live Voice in CodexGPT-Live FAQ
Is GPT-Live the same as Advanced Voice?
No. Live is the newest option powered by GPT-Live-1 or mini. Advanced is the previous realtime Voice mode and remains the choice for supported video and screen sharing.
Which model powers ChatGPT in the background?
OpenAI described GPT-5.5 as the background model for the July 8 consumer launch. For Voice in Work and Codex, current official pricing says GPT-Live handles the conversation while GPT-5.6 Terra starts and coordinates desktop tasks.
Can I build directly with GPT-Live-1?
Not through a generally available GPT-Live-1 API as of 2026-07-24. Use GPT-Realtime-2.1 or its mini version for current developer projects.