OpenAI launches GPT-Live voice model that listens while speaking
- Source
- OpenAI
- Time
- 5:13 PM
- Weight
- 95/100
OpenAI has launched GPT-Live, a new generation of voice models designed to facilitate more natural, human-like interactions with ChatGPT. Unlike previous systems that processed speech in discrete turns, GPT-Live utilizes a full-duplex architecture that allows it to listen and speak simultaneously.
This enables the AI to process interruptions, respond to brief acknowledgments like "mhmm," and maintain a continuous flow of conversation without the rigid pauses associated with older technology. The new system decouples conversational interaction from complex processing by delegating tasks such as web searches and deep reasoning to the company's latest frontier models, including GPT-5.5.