OpenAI has unveiled GPT-Live, a cutting-edge voice AI aimed at making interactions more fluid by enabling simultaneous speaking and listening. Utilizing a full-duplex architecture, GPT-Live can react to users’ speech in real-time, incorporating natural conversational cues like “mhmm” or “yeah,” which allows for more seamless and swift exchanges without the disruptions caused by lengthy pauses.
For tasks that demand more intricate processing, such as web searches or advanced reasoning, GPT-Live can effortlessly delegate to a more robust AI model operating in the background, all while maintaining the dialogue with the user. These sophisticated tasks are initially powered by GPT-5.5, with plans to integrate even newer models in upcoming updates.
The release of GPT-Live comes on the heels of OpenAI’s announcement that its GPT-5.6 model series, which includes the prominent Sol model as well as the Terra and Luna variants, will be available to the public once it passes an additional cybersecurity review. This new series represents the latest advancement in AI technology from OpenAI.
OpenAI has commenced the global rollout of two iterations of the new voice models—GPT-Live-1 and GPT-Live-1 mini—for ChatGPT users. Additionally, the company is preparing to offer GPT-Live through its API, which will enable developers and businesses to create applications with real-time voice AI functionalities.