Because of the complicated combination of mechanisms that combine to produce large power system failures, the simulation of cascading failure requires some modeling assumptions. In this paper we compare three models of cascading failure in electrical power systems under various assumptions. In the first, we combine dynamic generator models with a DC power flow network model, and use time-delayed (memory) relays to simulate branch failure. In the second, we simulate cascading with the use of sequential power flow calculations. In the third, we simulate cascading using a simple topological contagion model. The results indicate that the dynamical and the quasi-steady state (QSS) simulations show substantial agreement, whereas the topological model differs significantly. We also find that the extent of the agreement between the dynamical and the QSS model largely depends on the way in which branch failures occur.