>>512497170
So looks like the way it works is it just tries to guess to be the cheap/fast model or the more expensive reasoning model based off how easy the question looks. So it's like asking a question and sometimes a retard answers it and sometimes an expert answers it and you don't really know who will.