OpenAI has released GPT-Live, a new voice model designed to make interactions with ChatGPT more natural. This upgrade allows the AI to listen and speak simultaneously, incorporating acknowledgment phrases like "mhmm" to signal it's engaged. GPT-Live can also interrupt, pause, or decide to use tools multiple times per second, enabling continuous, fluid dialogue.

This model leverages OpenAI's latest frontier model, GPT–5.5, for advanced capabilities such as web search, complex reasoning, and task execution. While performing tasks, GPT-Live can maintain conversational flow and deliver results upon completion. It also supports live translation and displays visual information cards for weather, stocks, and sports.

GPT-Live–1 is now the default for ChatGPT Go, Plus, and Pro users, with a lighter version, GPT-Live–1 mini, available for Free users. OpenAI is actively developing support for voice with video and screen sharing, a feature not yet available. The introduction of GPT-Live signals a move towards more intuitive and responsive AI voice interfaces for consumers.