OpenAI has unveiled GPT-Live, a groundbreaking voice AI that enhances conversational interactions by enabling simultaneous speaking and listening. This innovation is built on a full-duplex framework, allowing the AI to interject with natural conversational cues like “mhmm” or “yeah,” thereby facilitating more fluid and rapid exchanges without the interruption of long pauses.
For tasks that demand more intricate processing, such as web searches or complex reasoning, GPT-Live efficiently delegates these to a more robust AI model operating in the background, all while maintaining the dialogue with the user. This capability is initially powered by GPT-5.5, with plans to incorporate newer, advanced models in subsequent updates.
The rollout of GPT-Live coincides with OpenAI’s announcement regarding the upcoming release of its GPT-5.6 model series. This new series, which includes the flagship Sol model along with Terra and Luna variants, is pending a final cybersecurity review before its public debut.
Two versions of the voice model, GPT-Live-1 and GPT-Live-1 mini, are being gradually introduced to ChatGPT users across the globe. Additionally, OpenAI intends to extend the reach of GPT-Live by offering it through an API, thus empowering developers and businesses to integrate real-time voice AI capabilities into their applications.