OpenAI has launched GPT-Live, a new set of voice AI models designed to make conversations with ChatGPT more natural while improving its ability to handle complex tasks in the background.
The technology uses a full-duplex architecture, allowing the model to listen and speak simultaneously rather than waiting for users to finish talking before responding. OpenAI said this enables more fluid conversations, with the assistant able to acknowledge users, pause when appropriate and support faster back-and-forth exchanges.
GPT-Live also separates conversational interactions from more demanding tasks such as web searches and advanced reasoning. When users ask more complex questions, the voice model can hand the request to OpenAI’s latest frontier model, GPT-5.5, while continuing the conversation until the response is ready.
The company said the approach allows users to access more advanced AI capabilities without interrupting the flow of the conversation.
OpenAI is rolling out two versions of the technology, GPT-Live-1 and GPT-Live-1 mini, to ChatGPT users globally across iOS, Android and the web. GPT-Live-1 will become the default voice model for Go, Plus and Pro subscribers, while Free users will receive GPT-Live-1 mini.
The update also introduces new features to ChatGPT Voice, including richer visual responses for topics such as weather, stocks and sport, alongside support for web search, memory, image uploads and file sharing.
According to OpenAI, more than 150 million people use ChatGPT’s voice and dictation features each week for tasks ranging from language learning to hands-free assistance.
The company added that it has expanded safety testing for GPT-Live to include audio-specific risks such as emotional reliance, self-harm, sexual content and voice misuse. It has also introduced safeguards designed to intervene during conversations if the system detects potentially unsafe responses.
OpenAI said GPT-Live currently supports its most widely used languages, although some users may notice non-native accents or reduced fluency in certain languages. Voice with video and screen sharing will not be available at launch but is planned for a future update.
