A proper class cannot be an element of a set or a class and is not subject to the Zermelo-Fraenkel axioms of set theory; thereby a number of paradoxes of naive set theory are avoided. Instead, these paradoxes become proofss that a certain class is proper. For example, Russell's paradox becomes a proof that the class of all sets is proper, and the Burali-Forti paradox becomes a proof that the class of all ordinal numbers is proper.
The standard Zermelo-Fraenkel set theory axioms do not talk about classes; classes exist only in the metalanguage as equivalence classes of logical formulas. Another approach is taken by the von Neumann-Bernays-Gödel axioms; classes are the basic objects in this theory, and a set is then defined to be a class that is an element of some other class. The proper classes, then, are those classes that are not elements of any other class.
Several objects in mathematics are too big for sets and need to be described with classes, for instance large categories or the class-field of surreal numbers.
The word "class" is sometimes used synonymously with "set", most notably in the term "equivalence class". This usage dates from a historical period where classes and sets were not distinguished as they are in modern terminology. Many discussions of "classes" in the 19th century and earlier are really referring to sets, or perhaps to a more ambiguous concept.