Why don't you simply ask the respective model which model is best for a specific task? :-)