OpenAI has launched GPT-Live, a new voice model for ChatGPT that makes conversations feel more natural. It has started rolling out on iOS, Android, and ChatGPT.com, giving users a smoother way to talk with AI.
GPT-Live can listen and speak at the same time. This means there are fewer pauses and more natural back-and-forth. It can even respond with short cues like “mhmm” or “got it.” For harder tasks, GPT-Live hands things off to GPT-5.5 or future models while keeping the chat going.
Older voice modes often felt slow or stiff. GPT-Live fixes that with faster responses and better flow. In tests, people preferred it over the Advanced Voice Mode. It also scored higher in reasoning, web search, and support tasks.
ChatGPT Voice now feels smarter and easier to use. You can adjust reasoning levels, pause or interrupt, and get visual cards for things like weather, stocks, and sports. It also handles background noise better. Safety is built in, with training for voice risks, real-time safeguards, and parental controls for teens.
GPT-Live comes in two versions: GPT-Live-1 for paid tiers (Go, Plus, Pro) and GPT-Live-1 mini for free users. It supports several languages. Video and screen sharing are not yet available but are planned.
Also Read: OpenAI breaks away from Microsoft, opens doors to rivals
With GPT-Live, ChatGPT takes a big step toward natural voice conversations. This sets up future updates that make AI feel even more like talking to a person.
