OpenAI has unveiled GPT-Live, a groundbreaking voice AI system designed to enhance the fluidity of conversations by enabling simultaneous speaking and listening. Utilizing a full-duplex architecture, GPT-Live can interject with natural responses like “mhmm” or “yeah,” promoting smoother interactions without the typical delays in communication.
For tasks that demand more intricate solutions, such as web searches or comprehensive reasoning, GPT-Live deftly delegates these to a more robust AI model operating in the background. At the time of its launch, these sophisticated functions are powered by the GPT-5.5 model. OpenAI plans to incorporate support for more advanced models in future iterations to further enhance capabilities.
The introduction of GPT-Live follows closely on the heels of OpenAI’s announcement regarding the upcoming public release of its GPT-5.6 series, pending final cybersecurity clearances. This series will include the flagship Sol model alongside the Terra and Luna variants, broadening the scope and application of OpenAI’s AI technology.
OpenAI has already begun the rollout of two versions of its new voice models—GPT-Live-1 and GPT-Live-1 mini—to ChatGPT users globally. The company is also preparing to offer GPT-Live through its API, which will allow developers and businesses to integrate real-time voice AI capabilities into their applications, expanding the potential use cases for this innovative technology.