GPT-Live (https://openai.com/index/introducing-gpt-live/) is OpenAI’s new voice model that powers ChatGPT Voice, real-time conversations with simultaneous listening and speech.

GPT-Live processes the audio input continuously while generating its response. It can interrupt, be interrupted, mark active listening (“mhmm”, “agree”) or remain silent (features also developed in Kyutai’s Unmute ).

For complex questions with web research or in-depth reasoning, GPT-Live delegates the background task to GPT-5.5 by holding the conversation, and then reintegrates the result when ready.

Visual responses displayed during the voice exchange: weather, stock price, sports scores, maps.

A feature built into ChatGPT, GPT-Live becomes the default voice model for paid accounts (Go, Plus, Pro), GPT-Live-1 mini for free accounts.

Accessible with a free ChatGPT account.