OpenAI today announced a major upgrade for ChatGPT Voice, powered by GPT‑Live, a new generation of voice models. This update marks the next major step for ChatGPT’s voice mode, which first reached a broader audience through Advanced Voice Mode in 2024.
Advanced Voice Mode was impressive when it launched, offering more natural conversations, faster responses, and support for interruptions. However, the underlying technology has quickly become outdated as OpenAI has moved to newer real-time audio models built specifically for more capable voice agents.
Back in May, OpenAI announced GPT-Realtime-2, its most advanced real-time voice model so far. Unlike earlier voice systems that mainly focused on fast speech-to-speech conversations, GPT-Realtime-2 brings GPT-5-class reasoning to live audio interactions. This allows voice apps built on this model to handle more complex requests, follow longer conversations more reliably, and carry out multi-step tasks while maintaining a natural speaking flow.
Today, OpenAI introduced the new ChatGPT Voice experience powered by GPT-Live, its new full-duplex voice model architecture. With this model, ChatGPT can speak, hear, and listen at the same time, instead of taking turns like a traditional voice assistant.
The model can naturally express small acknowledgments like “okay” or brief verbal nods like “mhmm” when the user pauses or slows down, without fully interrupting. It can also stay quiet when the user needs a moment to think. Users can interrupt at any time, and the model adjusts immediately.
GPT-Live can also answer questions that require web search, deeper reasoning, or more complex work, as it uses the latest frontier model behind the scenes to prepare the answer. For now, GPT-Live will use GPT-5.5 in the background.
OpenAI is rolling out two versions of GPT-Live, GPT-Live-1 and GPT-Live-1 mini, to ChatGPT users worldwide. ChatGPT Go, Plus, and Pro users will get the GPT-Live-1 model for their voice experience, while Free users will get a voice experience powered by GPT-Live-1 mini. These models are also coming to the API for developers and enterprises.