The cytochrome P450 enzymes collectively metabolize a wide range of xenobiotic and endogenous compounds. The broad substrate specificity of this superfamily derives from the multiplicity of P450s whose unique substrate specificity profiles reflect underlying differences in primary sequence. Experimental structures of P450s, where available, have provided great insight into the basis of substrate recognition. However, for those mammalian P450s whose structures have not been determined, homology modeling has become an increasingly important tool for understanding substrate specificity and mechanism. P450 modeling is often a challenging task, owing to the rather low sequence identity between target and template proteins. Although mammalian P450 models have previously been based on bacterial P450 structures, the recent advent of mammalian P450 structures holds great potential for generating more accurate homology models. Consequently, the substrate recognition properties of several mammalian P450s have been rationalized using the predicted substrate binding site of recently developed models. This review summarizes the major concepts and current approaches of molecular modeling of P450s.