Does auto mode use the same model in each turn? Or does it decide which model only before starting to answer?
Hey there!
Auto picks a model right before it answers each turn, and keeps that model for the whole of that answer. It decides fresh on every new message, so different turns in a conversation can use different models. It generally stays consistent from turn to turn, but that isn’t guaranteed. The choice can vary based on availability and current conditions.
Thanks for the reply!
So what happens if a question/task is actually much more difficult than it first seems to be? Doesn’t it end up with smaller models doing difficult tasks, which is not supposed to be? I’m just curious. 
For users worried about this, it makes sense to choose a named model rather than Auto!
Hahaha I still love using Auto and I’ll continue using it.
I just thought it might be better if there were ways for models to switch models by themselves during the thinking process.
Just out of interest, why is there no transparency around the model being used? What would the negative side of this be for Cursor or the user?