OpenAI GPT-Live: AI That Listens & Speaks Simultaneously

OpenAI has launched GPT-Live, a new generation of voice models for ChatGPT. This system can listen and speak simultaneously, making conversations feel much more natural. Here's the thing: instead of waiting for you to finish talking, GPT-Live engages in quick, back-and-forth exchanges. It can even use natural interjections like "mhmm" or "yeah" and knows when to pause. Two versions are rolling out: GPT-Live-1 for paid subscribers and GPT-Live-1 mini for free users. Both use GPT-5.5 for daily questions. For more complex tasks, the system delegates to the latest frontier model. What's interesting is this shifts AI conversations from a query-response cycle to a continuous stream. Developers and businesses can sign up for API notifications, opening doors for improved customer service and interactive workflows. The rollout is global, with the mini version for free users. This aims to make voice interaction the default experience. The bottom line: this new full-duplex approach sets a higher standard for natural AI conversation.