OpenAI has unveiled GPT-Live, its latest advancement in voice artificial intelligence technology, aimed at enhancing the fluidity and naturalness of spoken interactions. This innovation relies on a full-duplex system, enabling the AI to simultaneously speak and listen, thereby reducing conversational delays. It achieves this by employing natural conversational fillers like “mhmm” or “yeah,” which contribute to more seamless and rapid exchanges.
In situations that demand more intricate processing, such as conducting web searches or performing sophisticated reasoning, GPT-Live is engineered to delegate these tasks to a more robust AI model operating behind the scenes. This feature ensures the conversation continues uninterrupted. At the time of its launch, these complex tasks are executed by GPT-5.5, with plans to integrate support for emerging models in subsequent updates.
The introduction of GPT-Live coincides with OpenAI’s announcement regarding the upcoming public release of its GPT-5.6 model series. This release is contingent on the completion of additional cybersecurity evaluations. The GPT-5.6 lineup will include the flagship Sol model, as well as Terra and Luna variants.
OpenAI has already begun deploying two versions of these novel voice models, GPT-Live-1 and GPT-Live-1 mini, to users of ChatGPT across the globe. In addition to this, the company intends to extend the availability of GPT-Live through its API, which will allow developers and businesses to incorporate real-time voice AI capabilities into their applications.