ChatGPT's voice mode gets a new brain that can phone a bigger one
OpenAI's GPT-Live replaces the ChatGPT voice model and can hand harder questions to GPT-5.5 in the background, keeping the conversation going while it waits for the answer.
Buried under the GPT-5.6 noise, OpenAI also swapped the model behind ChatGPT’s voice mode. The new one is called GPT-Live, and its party trick is delegation: when a question needs web search or deeper reasoning, it hands the task to GPT-5.5 behind the scenes and keeps talking with you while the heavy model works, folding the answer back into the conversation when it lands. Simon Willison, who had preview access for a few weeks on the iPhone app, calls the result very impressive, and his track record on these calls is good.
The architecture is the takeaway: a small fast model holding the conversation while big models do the thinking is the pattern that makes voice assistants actually feel like assistants. It is also exactly the pattern local builders are converging on from the other end, small ASR and TTS models on CPU with a local LLM behind them. The frontier version and the bench version of this idea are closer than they look, and only one of them needs an account.