Diabetic nephropathy (DN) is a leading cause of end-stage renal disease in the developed world. Accordingly, an urgent need exists for new, curative treatments as well as for biomarkers to stratify risk of DN among individuals with diabetes mellitus. A barrier to progress in these areas has been a lack of animal models that faithfully replicate the main features of human DN. Such models could be used to define the pathogenesis, identify drug targets and test new therapies. Owing to their tractability for genetic manipulation, mice are widely used to model human diseases, including DN. Questions have been raised, however, about the general utility of mouse models in human drug discovery. Standard mouse models of diabetes typically manifest only modest kidney abnormalities, whereas accelerated models, induced by superimposing genetic stressors, recapitulate key features of human DN. Incorporation of systems biology approaches and emerging data from genomics and metabolomics studies should enable further model refinement. Here, we discuss the current status of mouse models for DN, their limitations and opportunities for improvement. We emphasize that future efforts should focus on generating robust models that reproduce the major clinical and molecular phenotypes of human DN.