OpenAI is rolling out new GPT-Live voice models to all ChatGPT users, featuring continuous interaction capabilities that allow the AI to listen, speak, and process information simultaneously.

Key Points

• The update introduces GPT-Live-1 and a mini version, enabling real-time, human-like conversation with filler words and reduced latency.
• A new continuous interaction framework allows the AI to listen and speak at the same time, facilitating live translation and multitasking.
• The system can delegate complex reasoning tasks to frontier models like GPT-5.5 while maintaining an active, seamless conversation with the user.
• ChatGPT can now generate visual outputs, such as weather reports or sports scores, to supplement spoken responses.
• The models are available to both free and paid users on mobile and web platforms, with a rollout period of up to two days.
• OpenAI has implemented expanded safety guardrails regarding self-harm, violence, and sexual content while automatically opting users out of voice data training.

Why it Matters

This update shifts AI interaction from a rigid, turn-based format to a fluid, human-like experience that mimics natural conversation. By enabling simultaneous processing and task delegation, OpenAI is positioning ChatGPT as a more versatile tool for real-time productivity and complex problem-solving.