ChatGPT Now Listens and Speaks Simultaneously

OpenAI just dropped GPT-Live, a new voice model that makes chatting with ChatGPT feel like talking to an actual person. Instead of the old awkward wait-your-turn setup, GPT-Live uses full-duplex tech—so it can listen and speak at the same time. It even throws in filler words like “mhmm” to show it’s paying attention. You can interrupt, pause, or ask it to slow down, and the model decides multiple times per second whether to talk, listen, or call a tool. There are two versions rolling out now: GPT-Live-1 for paying users and GPT-Live-1 mini for free accounts. For tough questions that need web searches or logical reasoning, GPT-Live hands those off to the more powerful GPT-5.5 in the background while keeping the convo flowing. Early tests show people preferred GPT-Live over the old Advanced Voice Mode by a wide margin. Why it matters: This shift from rigid turn-taking to fluid, interruptible dialogue marks a major leap toward natural human-AI interaction, potentially changing how we use AI for everyday tasks like quick research or casual brainstorming.