We use a global model of Earth's magnetosphere and ionosphere to simulate the Geospace Environment Modeling (GEM) substorm challenge event of November 24, 1996. We compare our results to International Monitor for Auroral Geomagnetic Effects (IMAGE) ground magnetometer data, assimilative mapping of ionospheric electrodynamics (AMIE) polar cap potential and field aligned current patterns, Polar Visible Imaging System (VIS) estimates of the polar cap magnetic flux, GOES 8 geosynchronous magnetometer data, IMP 8 magnetometer data, and Geotail plasma and magnetic field data. We find generally good agreement between the simulation and the data. The modeled evolution of this substorm generally follows the phenomenological near-Earth neutral line model. However, reconnection in the tail is very localized, which makes establishing a causal relation between tail dynamics and auroral dynamics difficult, if not impossible. We also find that the model results critically depend on the parameterization of auroral Hall and Pedersen conductances and anomalous resistivity in the magnetosphere. For many combinations of parameters that enter these parameterizations, no substorm develops in the model, but instead the magnetosphere enters a steady convection mode. The main deviation of the model from the data is excessive convection, which leads to a strong, driven westward electrojet in the growth phase, only partial tail loading, and a reduced recovery phase. Possible remedies are a better model for auroral conductances, an improved anomalous resistivity model, and a more realistic treatment of the ring current. ¿ 2001 American Geophysical Union