OpenAI has updated ChatGPT‘s voice mode with a new model called GPT-Live-1 that reduces interruptions and waits for users to finish speaking before responding. This change addresses the common frustration of conversational AI cutting off speakers mid-sentence or launching into monologues without checking for input. The system now routes complex queries to stronger text models like GPT-5.5 when reasoning or searching the web is required, allowing faster transitions between research and dialogue. Kundan Kumar, the lead researcher behind the project, described this as the company’s most intelligent voice model to date. The update aims to make interactions feel more like natural conversations between people rather than a rigid question-and-answer session. By improving how the software handles pauses and manages information retrieval, OpenAI is refining the practical usability of its voice interface for daily tasks.
- GPT-Live-1 automatically pauses to let users finish their thoughts.
- Complex questions are routed to GPT-5.5 for better reasoning.
- The interface now feels more like talking to another person.
