We present a short overview of developments of the last decade in asymptotic analysis of extrema of families of random variables. We focus on the methods of investigating the quality of approximations as given by Gnedenko’s extreme value theorem, and its generalizations to the case of dependent random variables.