A grease filter is used to remove grease generated from a cooking appliance in a kitchen. Since the inertial impaction is a dominant particle removal mechanism of the grease filter, the performance of the filter is greatly affected by the geometry. This numerical study has been conducted to investigate the effect of geometry on the performance of grease filters for four models having nominal flowrate of 100 m$^3$/h. Four models were designed by changing the shape of impaction surface, the length of eyelid, and the number of eyelids of the grease filter. The flow field and particle trajectories in the grease filter with a flow chamber were simulated using the commercial code of STAR-CD. The difference of air velocity and pressure distributions among four models was discussed in detail. The collection efficiency curves and the pressure drops of four models were also compared. It was found that the grease filter model with flat top surfaces shows highest performance among four models, having high particle collection efficiency and relatively low pressure drop. The cutoff diameter of this model representing 50-% collection efficiency is about 7.1 ${mu}{ extrm}{m}$ for water droplets at 100 m$^3$/h.