Introducing GPT-Live

What's new

• GPT-Live-1: flagship full-duplex voice model — listens and speaks simultaneously for natural conversation
• GPT-Live-1 mini: default voice model in ChatGPT, replacing Advanced Voice Mode
• Natural interruptions: responds to interruptions, inserts 'mhmm', and stays quiet when user is thinking
• Delegation to GPT-5.5: for deep reasoning or web search, GPT-Live hands off to GPT-5.5 in the background
• Live translation: real-time simultaneous translation during conversation
• Visual cards: weather, stocks, sports scores on relevant queries
• API access: coming later for developers via signup form

Why it matters

GPT-Live replaces Advanced Voice Mode for all ChatGPT users globally. Full-duplex architecture fundamentally changes voice AI UX — conversation feels like a phone call. API access will open new voice assistant possibilities in mobile apps.

How to try it

Available in ChatGPT on iOS and Android; GPT-Live-1 mini is the new default voice model. API access: signup form on openai.com (date unannounced).