OpenAI launched GPT-Live, a new generation of voice models designed to make conversations with AI more natural while allowing ChatGPT to handle tasks during a voice conversation. The technology powers ChatGPT Voice and enables it to listen and speak at the same time.
GPT-Live is built around a full-duplex voice architecture, allowing the model to listen while speaking. It can respond to pauses, interruptions and changes in the direction of a conversation rather than waiting for the user to finish speaking before responding.
OpenAI says the model can make interaction decisions many times each second, including whether to speak, continue listening, pause, interrupt or invoke a tool. It can also use brief responses such as ‘mhmm’ or ‘yeah’ while listening, helping conversations feel less rigid.
OpenAI has introduced two versions, GPT-Live-1 and GPT-Live-1 mini, which are being rolled out to ChatGPT users globally. At launch, GPT-Live uses GPT-5.5 in the background when a request requires deeper reasoning, web search or more complex work.
The update also expands what users can do through Voice. On the web and mobile, users can ask Voice to work with supported plugins and connected apps available in ChatGPT. Existing permissions and usage restrictions continue to apply.
For users with access to ChatGPT Work, Voice can create documents, presentations and spreadsheets, work with connected apps and use a browser. If a user ends a voice conversation while a Work task is still running, the task can continue in text.
OpenAI is also bringing Voice to its ChatGPT desktop experience on macOS and Windows. Users can use spoken instructions to start tasks, check progress, ask questions about agents and coordinate multiple agents.
The feature is rolling out to Plus, Pro, Business, Edu and Enterprise plans. Users can also direct agents such as ChatGPT Work and Codex through voice.
GPT-Live also connects spoken conversations with regular ChatGPT chats. Voice responses can appear as rich text, allowing users to review information after speaking and switch between voice and typing.
Some examples of applications include preparation for the day, prioritizing, brainstorming, document creation, trip planning and shopping recommendations. GPT-Live-1 and GPT-Live-1 mini have been integrated into ChatGPT Voice but vary with regard to their availability and usage limits depending on the plan.
Also Read: ChatGPT Gets Real-Time Voice Upgrade with GPT-Live Rollout