OpenAI has unveiled its latest innovation in voice artificial intelligence, GPT-Live, which aims to enhance the natural flow of conversations by enabling the AI to speak and listen simultaneously. This advancement is made possible through a full-duplex architecture, allowing GPT-Live to interject with natural acknowledgments like “mhmm” or “yeah” while users are still speaking. The new system promises smoother and quicker interactions without the interruptions caused by lengthy pauses.
For tasks that require more complex processing, such as web searches or intricate reasoning, GPT-Live can efficiently delegate these challenges to a more powerful AI model operating in the background. Currently, these sophisticated tasks are managed by GPT-5.5, with plans to integrate support for newer, cutting-edge models in forthcoming updates.
This development coincides with OpenAI’s announcement regarding the upcoming public release of its GPT-5.6 model series. The release is contingent upon passing a comprehensive cybersecurity review. The GPT-5.6 lineup features the flagship Sol model, in addition to the Terra and Luna variants, highlighting OpenAI’s commitment to advancing its AI capabilities.
OpenAI is progressively deploying two versions of its new voice models—GPT-Live-1 and GPT-Live-1 mini—to users of ChatGPT around the globe. In addition, the company intends to extend the availability of GPT-Live through its API, thereby allowing developers and businesses to incorporate real-time voice AI capabilities into their applications.