AI Pulse by Inblix

GPT-Live: Real-Time Voice AI That Listens & Talks Back

OpenAI Blog · Jul 9, 2026 · 1 min read · Read original article →

Curated by the Inblix editorial team


OpenAI just dropped GPT-Live, a next-gen voice model that changes how we chat with AI. Unlike older systems that forced you to wait for a turn, GPT-Live uses a full-duplex architecture, meaning it can listen and speak simultaneously. It’ll throw in a “mhmm” or stay quiet when you’re thinking, making conversations feel natural. It’s also smart—when you ask for complex stuff like web searches or reasoning, it secretly hands off to a frontier model like GPT-5.5 while maintaining the chat flow. Two versions are rolling out globally today: GPT-Live-1 and GPT-Live-1 mini, with an API coming soon. So, forget the clunky, turn-based chat bots of the past. GPT-Live aims to make human-AI collaboration as fluid as talking to a person. Why it matters: This shifts voice AI from a rigid tool to a conversational partner, unlocking potential for more agentic, longer-running tasks and redefining how we interact with machines.

💡 Key Takeaways

  1. GPT-Live can listen and speak at the same time with full-duplex technology, allowing for interruptions and natural pauses.
  2. It delegates complex tasks like web searches to a frontier model (currently GPT-5.5) while keeping the conversation flowing seamlessly.
  3. Two versions launched today for ChatGPT users, with API access available for developers soon.

Keep reading: See related articles below for more coverage on this topic.

Get smarter about AI

The sharpest AI news, curated daily. Delivered free to your inbox.

Learn more

Glossary terms

← Back to all articles