Source: OpenAI — 2026-07-08
Summary
OpenAI launched GPT-Live, a new generation of voice models built on a full-duplex architecture — meaning the model can listen and speak at the same time, rather than taking strict conversational turns. Two variants shipped: GPT-Live-1 and GPT-Live-1 mini, rolling out globally to power ChatGPT Voice. At launch, GPT-Live runs on GPT-5.5 in the background for reasoning, with the voice layer handling natural conversational behaviors like backchannel acknowledgments ("mhmm," "yeah") and knowing when to stay quiet.
Key Takeaways
- Full-duplex architecture lets GPT-Live listen and speak simultaneously, enabling more natural back-and-forth and interruption handling than turn-based voice models.
- GPT-Live-1 becomes the default voice model for ChatGPT Go, Plus, and Pro subscribers; GPT-Live-1 mini becomes the default for Free-tier users.
- Runs on GPT-5.5 as the underlying reasoning model at launch.
- OpenAI cites more than 150 million weekly users of ChatGPT's Voice and Dictation features, framing this as a significant upgrade to an already heavily-used surface.