A good representation model of tolerance information is indispensable for achieving computer-aided tolerancing. Currently, the representation model used in industry is the EXPRESS model, while at the same time, a number of other kinds of representation models have been presented within the academia. The coexistence of different kinds of representation models generates a series of questions and discussions: can the EXPRESS model completely implement the representation of tolerance information semantics in an explicit, computer-readable, and computer-interpretable way? What challenges have been addressed to date by the presented representation models? What are the advantages and disadvantages of each representation model? What capabilities should an ideal representation model have? What are the potential research directions in tolerance information representation in the future? To approach these questions, a review of representation models of tolerance information is presented in this paper. An in-depth analysis of existing representation models is firstly provided. Then, the review makes a detailed comparison among them based on this analysis. Finally, some future research directions in tolerance information representation are suggested.