Comment by don_searchcraft

Pretty interesting to see how widely the token use differs across the models for the same task.