%20--%3e%3cdefs%3e%3cstyle%3e%20.st0%20{%20fill:%20%23061b40;%20}%20.st1%20{%20fill:%20%23306af1;%20}%20.st2%20{%20fill:%20%235ce5cf;%20}%20%3c/style%3e%3c/defs%3e%3cg%3e%3cpath%20class='st0'%20d='M55,10.5h9v3h-9v9h12V7.5h-12v3ZM64,19.5h-6v-3h6v3Z'/%3e%3cpolygon%20class='st0'%20points='69%2016.5%2078%2016.5%2078%2019.5%2069%2019.5%2069%2022.5%2081%2022.5%2081%2013.5%2072%2013.5%2072%2010.5%2081%2010.5%2081%207.5%2069%207.5%2069%2016.5'/%3e%3cpolygon%20class='st0'%20points='95%2010.5%2095%207.5%2083%207.5%2083%2022.5%2095%2022.5%2095%2019.5%2086%2019.5%2086%2016.5%2095%2016.5%2095%2013.5%2086%2013.5%2086%2010.5%2095%2010.5'/%3e%3cpath%20class='st0'%20d='M40,1.5v21h11.6l1.4-1.4v-7.6h0c0,0-1.4-1.5-1.4-1.5l1.4-1.4V2.9l-1.4-1.4h-11.6ZM50,19.5h-7v-6h7v6ZM50,10.5h-7v-6h7v6Z'/%3e%3c/g%3e%3cpath%20class='st1'%20d='M23.1,24L14.7,4.8l-4.9,11.2h3.8l-1.8,4H3.3L12.1,0H2C.9,0,0,.9,0,2v20c0,1.1.9,2,2,2h21.1Z'/%3e%3cpath%20class='st2'%20d='M34,0h-16.8l10.6,24h6.2c1.1,0,2-.9,2-2V2C36,.9,35.1,0,34,0ZM32.5,20h-4V4h4v16Z'/%3e%3c/svg%3e)
OpenAI officially launched a new ChatGPT voice experience based on the GPT-Live architecture, marking a major architectural leap in its voice technology. Compared to the Advanced Voice Mode launched in 2024, the new version no longer relies on traditional voice systems but instead adopts a real-time audio model specifically designed for intelligent voice agents.
As early as May this year, OpenAI announced the most advanced real-time speech model, GPT-Realtime-2, which introduces GPT-5 level reasoning capabilities into real-time audio interaction scenarios. This model can better handle complex requests, track long conversations more stably, and complete multi-step tasks while maintaining natural speech speed and tone. The new voice experience based on the GPT-Live architecture now further raises the interaction limit, allowing ChatGPT to truly have the ability to "speak, listen, and think at the same time."
It can respond like a human, without interrupting, and even use "okay" or "uh-huh."
The full-duplex architecture is the core of this upgrade, breaking the traditional voice assistant's "you speak first, then I answer" queue mode. In actual interactions, GPT-Live can naturally insert slight intonation feedback such as "okay" or "uh-huh" when the user pauses or slows down their speech, without abruptly interrupting the complete sentence. When the user needs a moment to think, the model will remain silent actively, avoiding any disturbance.
In terms of interruption control, it also performs excellently. Users can interject or change the topic at any time, and the model will immediately adjust the response logic to smoothly connect with the new conversation goal. More importantly, the new voice experience can also handle questions requiring online search, deep reasoning, or complex task planning. By default, the background uses the GPT-5.5 model to generate answers. This means that users receive not only a dialogue rhythm closer to humans in voice scenarios but also the "deep thinking" capability similar to the text mode.
Two versions are released simultaneously, covering all users from free users to enterprise customers.
OpenAI also launched two versions, GPT-Live-1 and GPT-Live-1 mini, to cover different levels of users. Paid users such as ChatGPT Go, Plus, and Pro default to using the more comprehensive GPT-Live-1, while free users get a voice experience driven by the mini version. For developers and enterprise customers, both models are accessible through APIs, making it convenient for them to build next-generation voice agents and vertical applications.
Looking at the pace, OpenAI is accelerating the upgrade of the "walkie-talkie-style" voice assistant into a full-duplex intelligent agent model that is more similar to human conversation. With the launch of GPT-Realtime-2 and GPT-Live, the ChatGPT voice mode is evolving from a simple voice Q&A tool into an "intelligent conversation partner" capable of performing complex tasks and maintaining long-term interactions in voice environments.