July 9 News.OpenAI Official launch of the new-generation voice model GPT-Live, for iOS, Android and ChatGPT.com users worldwide。
The central change in GPT-Live is the architecture upgrade: from a "turn system" that used to rely on silent sound testing to a full-scale structure that allows both voice input and output。
instead of waiting for the user to speak, the model can give immediate feedback, such as "mmm" "got it" when the user speaks, and can wait when the user stops, without interrupting the conversation due to environmental noise errors。
In dealing with complex issues, GPT-Live gives the task backstage to GPT-5.5 to complete the search or reasoning, while maintaining a normal dialogue at the front end without the user feeling a clear wait。
Official tests show that GPT-Live-1 has an overall preference of 75.71 TP3T, which is significantly better than the advanced voice pattern of the previous generation, in terms of the dialogue experience. In addition, a rich media card has been added to the dialogue interface, and information on weather, stocks, etc. can be shown in synchronised text during the conversation。