OpenAI’s GPT-Live brings full-duplex voice to ChatGPT

OpenAI has introduced GPT-Live-1 and GPT-Live-1 mini, a pair of voice models designed to make spoken conversations with ChatGPT feel more natural. The headline change is full duplex: the models can listen and speak at the same time, so interruptions and turn-taking no longer require waiting for the other side to finish.

ChatGPT Voice now runs on GPT-Live-1 for paid users and GPT-Live-1 mini for free users. The models work inside a normal ChatGPT chat, with spoken replies appearing alongside streamed text, and can draw on web search and memory, show visual results through supported widgets, and handle text and images in the same exchange.

Alongside the consumer rollout, OpenAI released gpt-realtime-2.1 and gpt-realtime-2.1-mini for developers building low-latency voice and multimodal apps. The company reports at least 25% lower p95 latency across its realtime voice models, plus better speech recognition, noise handling, tool use and instruction following.

OpenAI also says the launch adds stronger voice safety safeguards and broader red-teaming before release.

Sources: OpenAI Newsroom; Build Fast with AI.

Drafted with AI assistance and reviewed by the IT Magazine news desk. Facts verified against the linked sources. Spotted an error? See our Corrections Policy.