OpenAI Launches GPT-Live Full-Duplex Voice Model

OpenAI has unveiled GPT-Live, the latest generation of voice models, now integrated into ChatGPT's "Conversation" mode. The key feature is its full-duplex architecture, enabling the neural network to listen and speak simultaneously without waiting for silence. The model understands interjections, can pause appropriately, and does not interrupt users who are thinking.

The developers have separated the conversation process from task execution. While GPT-Live handles the dialogue, another model (initially GPT-5.5) works in the background to search the internet or perform complex calculations. This allows real-time speech translation and solving requests without interrupting the conversation.