A critical review of published thrust models for applied-field magnetoplasmadynamic thrusters is presented, along with a new model addressing shortcomings related to electrode and magnet geometry. While numerous theoretical thrust models have been presented in the literature, there has not been a comprehensive comparison to determine which best predicts thruster behavior across a large parameter space. In order to make this determination, all thrust data available in the literature were collected into a single database and tested against each model. Unlike previous comparisons between prediction and measurement, only the regime in which the applied-field thrust component dominates is examined, allowing for a direct comparison without invoking models of self-field and gasdynamic components of thrust. The degree to which each model deviates from measurement is determined for each controllable parameter. It is found that the largest deviations are due to incorrect representation of the effects of electrode and solenoid geometries. In light of this comparative study, an improved empirical model is derived as a function of nondimensional parameters representing these geometric variables. The improved agreement is attributed to the effects of magnetic field topology near the anode, which sets the effective anode radius that controls the magnitude of the Lorentz force for certain electrode geometries.