OpenAI introduces GPT Live, a new voice model that upgrades ChatGPT’s voice experience with more natural, real-time conversation. OpenAI says GPT Live powers ChatGPT Voice so the assistant can listen and speak at the same time, using a “full duplex” approach rather than the earlier turn-based setup that required pauses between user speech and the AI reply. Sources describe improved handling of interruptions and pauses, including more fluid back-and-forth timing and the ability for users to “cut in” without awkward delays.
Several outlets also say GPT Live improves performance on harder requests. When a query needs deeper reasoning or other work such as web search, GPT Live can delegate the task to a larger model in the background while the conversation continues. At launch, this background model is reported as GPT 5.5.
OpenAI’s update also adds or enhances features tied to the voice experience, including live translation, support for visual response cards (such as weather or stock information) during voice interactions, and improved noise handling during listening. GPT Live rolls out globally across iOS, Android, web, and is also referenced for CarPlay. It is offered in two versions—GPT Live 1 for paid tiers and GPT Live 1 mini for free users—with plans to bring the technology to the API soon.
