OpenAI Launches GPT-Live to Revolutionize ChatGPT Voice
OpenAI has officially launched GPT-Live, a groundbreaking new family of voice models designed to make talking to an artificial intelligence feel exactly like a natural human conversation. Rolled out globally on Wednesday, July 8, this new architecture entirely replaces the previous ChatGPT Advanced Voice Mode.
The IPO-bound AI startup introduced two versions of the model: GPT-Live-1 and the lightweight GPT-Live-1 mini. Both versions are now rolling out to ChatGPT users across iOS, Android, and the web.
A Full-Duplex Conversational Experience
In the past, interacting with voice assistants required a strict turn-based approach: you speak, wait for the AI to process, and then listen to the response. GPT-Live completely eliminates this delay by utilizing a "full-duplex" architecture.
Simultaneous Processing: GPT-Live can listen and speak at the exact same time.
Natural Interruptions: Users can easily interrupt the AI mid-sentence with a new thought, pause to gather their words, or ask the assistant to slow down.
Active Listening: To show it is paying attention, the model will interject with natural human acknowledgments like "mhmm," "yeah," or "got it" while you are still speaking.
Delegating Complex Work to GPT-5.5
While maintaining a fluid, uninterrupted conversation, GPT-Live is capable of handling complex background tasks. If a user asks a difficult question that requires deep reasoning or current internet data, the voice model seamlessly delegates the heavy lifting to OpenAI's latest frontier model, GPT-5.5.
While GPT-5.5 searches the web or processes complex logic, GPT-Live keeps the conversation flowing naturally until the task is finished. The update also brings visual aids directly into the voice interface, displaying rich visual cards on your screen for real-time information like weather forecasts, stock prices, and sports scores.
Additionally, the new architecture significantly improves real-time language translation, keeping pace with a speaker naturally as a conversation unfolds rather than waiting for a complete pause.
Availability and Subscription Tiers
OpenAI is rolling out the new experience across its entire user base, but access to the flagship model is gated by subscription tiers:
GPT-Live-1: The default, high-performance model for all ChatGPT Go, Plus, and Pro subscribers.
GPT-Live-1 mini: The default fallback model for Free tier users.
Currently, GPT-Live does not support voice interactions paired with live video or screen sharing, though OpenAI has stated that they are actively working to add those capabilities in the near future.