Introducing GPT-Live
OpenAI Blog · 2026-07-08
OpenAI launches GPT-Live, a full-duplex voice AI model that listens and speaks simultaneously, delegates complex queries to GPT-5.5 running in the background, and outperforms Advanced Voice Mode on conversational naturalness, expert scientific reasoning, and agentic web search benchmarks.
Extraction
Topics: voice-aifull-duplex-architecturechatgptmultimodal-aiai-safety
Claims
- GPT-Live uses a full-duplex architecture that continuously processes input while generating output, enabling natural back-and-forth, active listening cues, and real-time interruption handling without discrete turn-taking.
- GPT-Live delegates tasks requiring web search, deeper reasoning, or agentic capabilities to GPT-5.5 running in the background while maintaining uninterrupted conversational flow.
- GPT-Live-1 substantially outperforms Advanced Voice Mode on GPQA expert-level scientific reasoning and shows strong gains on BrowseComp agentic web search.
- More than 150 million people per week use ChatGPT voice features including Voice and Dictation.
- GPT-Live includes built-in real-time safeguards that can steer, modify, or end voice conversations when potentially unsafe output is detected, with dedicated protections for teen users.
Key quotes
GPT‑Live is built on a full-duplex architecture, meaning it can listen and speak at the same time.
Our vision is to enable truly natural human–AI interaction: a world where collaborating with AI feels as fluid and responsive as working with another person, while reasoning and complex task execution happen seamlessly in the background.
Over time, we believe this research will also unlock the ability to use voice for increasingly complex, longer-running, and more agentic work.