OpenAI released GPT-Live, a new generation of voice models including GPT-Live-1 and GPT-Live-1 mini. These models utilize a full-duplex architecture to listen and speak simultaneously, replacing previous turn-based systems that required users to finish speaking before receiving a response.
The continuous interaction framework eliminates pauses and allows for natural interruptions to make conversations more fluid. These models now power ChatGPT Voice and are rolling out globally to all users.
GPT-Live delegates complex queries and web searches to background models like GPT-5.5 and integrates the results into the conversation. The technology enables new capabilities such as simultaneous live translation, positioning OpenAI to compete with voice assistants from Google, Apple, and Amazon.