OpenAI launches GPT-Live voice model for simultaneous listening and speaking in ChatGPT

OpenAI has launched GPT-Live, a new generation of full-duplex voice model that enables ChatGPT to listen and speak simultaneously, transforming AI interactions into fluid, real-time conversations that feel more natural. This smartest voice model yet can demonstrate active listening with phrases like "mm" or "yeah," handle overlapping speech, and seamlessly delegate complex tasks like web searches or deep reasoning to its underlying frontier model (GPT-5.5 at launch) without breaking the flow of conversation. GPT-Live is rolling out globally today to ChatGPT users in two versions, GPT-Live-1 and GPT-Live-1 mini, with API access for developers and enterprises planned soon, marking a shift from simple call-and-response to voice agents capable of listening, reasoning, translating, and taking action as conversations unfold. The model is designed strictly for conversation with predefined voices and safeguards to prevent impersonating real people, making it a tool for genuine human-AI communication rather than voice cloning. Users can immediately tap the Voice button in ChatGPT to experience improved listening, smarter answers, and visual responses powered by this technology.