GPT-Live
OpenAI's real-time voice model family that listens and speaks at the same time, powering natural full-duplex conversations inside ChatGPT.
Updated 2026-07-09
Overview
GPT-Live is a family of real-time voice models OpenAI launched on July 8, 2026, designed so the assistant can listen and speak simultaneously rather than waiting for you to finish before it responds. That "full-duplex" behavior is the headline difference: you can interrupt mid-sentence, talk over the model, or trade quick back-and-forth without the turn-taking lag that made earlier voice assistants feel like walkie-talkies. It ships directly inside ChatGPT and rolled out to users worldwide at launch, including the free tier.
It's aimed at anyone who wants to actually converse with an AI out loud — hands-free brainstorming, language practice, cooking or driving with a running dialogue, or accessibility use cases where typing is a barrier. Because it lives in ChatGPT rather than as a separate app, there's nothing new to install or subscribe to; if you already use ChatGPT's voice mode, GPT-Live is the engine underneath it.
What sets it apart from the crowd of voice-agent platforms (Retell, Vapi, Bland, Hume) is that those are developer tools for building your own phone and support agents, while GPT-Live is a consumer-facing conversation layer wired into the most widely used chatbot on the market. The trade-off is the flip side of that convenience: it's not a standalone product with its own API story or pricing — you get it on OpenAI's terms, inside OpenAI's app.
Key features
Full-duplex conversation
The model can listen and speak at the same time, so you can interrupt or talk over it and it adjusts in real time instead of finishing a scripted response first.
Native ChatGPT integration
GPT-Live powers voice mode inside the existing ChatGPT apps, so there's no separate download, login, or subscription to manage.
Worldwide day-one rollout
OpenAI pushed it to ChatGPT users globally at launch, including the free tier, rather than staging it behind a waitlist or Pro-only gate.
Low-latency turn-taking
Simultaneous listening and speaking removes the pause-and-wait rhythm of older voice assistants, making spoken exchanges feel closer to talking to a person.
Pricing
Free tier: Available to ChatGPT free-tier users at launch, subject to usage limits
| Plan | Price | What's included |
|---|---|---|
| Free | $0 | Access to GPT-Live voice conversations in ChatGPT with standard usage limits |
| ChatGPT Plus | $20/mo | Higher usage limits and priority access to newer models and features |
| ChatGPT Pro / Team / Enterprise | From $200/mo (Pro) | Highest limits and org features — check OpenAI's pricing page for current tiers |
Access to GPT-Live voice conversations in ChatGPT with standard usage limits
Higher usage limits and priority access to newer models and features
Highest limits and org features — check OpenAI's pricing page for current tiers
Pros & cons
Pros
- ✓Full-duplex voice lets you interrupt and talk over the model naturally, unlike turn-based assistants
- ✓Available on the ChatGPT free tier from day one — no waitlist or paywall to try it
- ✓Nothing new to install or subscribe to if you already use ChatGPT
- ✓Global rollout at launch rather than a staged US-first release
Cons
- ×Not a standalone product — you only get it inside ChatGPT, on OpenAI's terms
- ×No dedicated developer API story like Vapi, Retell, or Bland for building your own voice agents
- ×Free-tier access comes with usage limits OpenAI hasn't fully detailed publicly
- ×Very new (launched July 2026), so real-world reliability across accents, noise, and edge cases is still unproven
How it compares
| Tool | Best for | Pricing | Score |
|---|---|---|---|
| GPT-Live | — | Free tier + ChatGPT Plus $20/mo | 8.7/10 |
| ChatGPT | — | Free tier + Plus $20/mo + Pro $200/mo | 9.5/10 |
| Claude vs Claude → | — | Free tier + Pro $20/mo + Team $30/mo/user | 9.5/10 |
| Gemini vs Gemini → | — | Free tier + Advanced $19.99/mo | 9.2/10 |
Compare head-to-head
Related reading
GPT-Live: OpenAI's Real-Time Voice Models Explained
OpenAI launched GPT-Live and GPT-Live mini on July 8, 2026 — voice models that listen and speak at the same time. Here's what shipped.
GPT-5.6 Sol Public Launch: What Changes Thursday
OpenAI moves GPT-5.6 Sol, Terra, and Luna from limited preview to broad public release Thursday after US approval. What actually changes.
Mistral Robostral Navigate: One Camera, 8B Params
Mistral's Robostral Navigate is an 8B robotics navigation model that runs on a single RGB camera and claims 76.6% on R2R-CE.
Ready to try GPT-Live?
Head to the official site to start with GPT-Live — pricing and plans are listed above.
Visit GPT-Live