On July 9, 2026 (Beijing Time), OpenAI rolled out a new voice model lineup, GPT-Live-1 and GPT-Live-1 mini, marking a significant leap forward in ChatGPT's voice capabilities. These innovative models incorporate a full-duplex architecture, facilitating seamless, simultaneous listening and speaking. This advancement allows users to interrupt the AI's responses effortlessly, fostering a more fluid and natural interaction experience. GPT-Live-1 is tailored for Go, Plus, and Pro subscribers who have opted for paid plans, whereas GPT-Live-1 mini is accessible to users on the free tier. The new models excel in accurately discerning when a user pauses or completes their statement, dynamically adjusting speech speed, tone, and emotional resonance in real-time. Furthermore, they can tap into GPT-5.5's capabilities in the background for intricate reasoning tasks or internet searches, presenting users with up-to-the-minute information such as weather forecasts and stock market updates. Notably, the new models also introduce enhanced voice safety protocols, automatically terminating conversations in high-risk scenarios. Additionally, parents have the option to disable voice functionality for minors or opt to receive alerts regarding potentially risky content.