OpenAI has upgraded the model powering ChatGPT's voice mode, introducing a new system called GPT-Live that utilizes more recent capabilities than the previous GPT-4o-based implementation.

  • The new GPT-Live system delegates complex tasks requiring web search or deeper reasoning to GPT-5.5 in the background while maintaining conversation flow.
  • At launch, GPT-Live uses GPT-5.5 as its backend frontier model, with plans to continuously update it as new models are released.
  • The previous voice mode relied on a GPT-4o era model with a knowledge cutoff from 2024, which the author found limited for brainstorming.

This update addresses the limitations of the older model, allowing for more useful interactions and handling harder tasks without interrupting the user's experience.