OpenAI Launches GPT-Live-1, a Full-Duplex Voice Model for ChatGPT
OpenAI replaced Advanced Voice Mode with GPT-Live-1 and a free-tier GPT-Live-1 mini, full-duplex models that can listen and speak at once and delegate hard reasoning to GPT-5.5.
OpenAI shipped GPT-Live-1 and GPT-Live-1 mini on July 8, retiring Advanced Voice Mode in favor of a full-duplex architecture that listens and speaks at the same time rather than waiting for turns to end. The models decide multiple times per second whether to speak, pause, interrupt, or hand off to a tool, and offload harder reasoning or web search to GPT-5.5 while keeping the conversation flowing.
- GPQA scientific reasoning: 84.2% for GPT-Live-1 at High reasoning, up from 45.3% for Advanced Voice Mode
- BrowseComp web-search task completion: 75.2%, versus 0.7% for the predecessor
- Human preference over Advanced Voice Mode: 75.7% for GPT-Live-1, 69.2% for GPT-Live-1 mini
- GPT-Live-1 is now default voice for ChatGPT Go, Plus and Pro; GPT-Live-1 mini serves free accounts; API access is coming soon
The rollout landed on iOS, Android and ChatGPT.com worldwide, with nine redesigned voices and visual cards for weather, stocks and sports scores; video and screen sharing are planned next. The launch lands two days ahead of GPT-5.6 Sol's broader general availability, giving OpenAI two separate model headlines in the same week.