OpenAI has launched GPT-Live, a cutting-edge voice AI system engineered to enhance the fluidity of conversations by allowing simultaneous speaking and listening. This new AI model is based on a full-duplex architecture, which enables it to respond to users even as they are speaking. It does so by incorporating natural fillers like “mhmm” or “yeah,” facilitating more seamless and quicker interactions without the awkward pauses that often characterize human-AI exchanges.
When it comes to handling more intricate tasks that involve web searches or require sophisticated reasoning, GPT-Live is designed to delegate these tasks to a more robust AI model working in the background, all while maintaining the flow of conversation with the user. Currently, these complex tasks are managed by GPT-5.5, with plans to support even more advanced models in future updates.
This development comes on the heels of OpenAI’s announcement that its premier GPT-5.6 model series will soon be available to the public, pending the completion of additional cybersecurity evaluations. The GPT-5.6 lineup features the top-tier Sol model, as well as the Terra and Luna variants, which are expected to offer a range of new capabilities.
In an effort to broaden the accessibility of its new voice models, OpenAI has already started the rollout of two versions—GPT-Live-1 and GPT-Live-1 mini—to ChatGPT users around the globe. Furthermore, the company has plans to extend GPT-Live’s reach by offering it through an API, which will empower developers and businesses to incorporate real-time voice AI functionalities into their own applications.