NASH arises from a variable interplay between environmental factors and genetic determinants that cannot be completely replicated in animals. Notwithstanding, preclinical models are needed to understand NASH pathophysiology and test mechanism-based therapies. Among several mouse models, some exhibit the key pathophysiologic as well as histopathologic criteria for human NASH, whereas others may be useful to address specific questions. Models based on over-nutrition with adipose restriction/inflammation and metabolic complications, particularly insulin resistance may be most useful to interrogate critical etiopathogenic factors. In-depth pathologic description is required for all models. Some demonstrate hepatocyte ballooning, which can be confused with microvesicular steatosis, while demonstration of an inflammatory infiltrate and pattern of liver fibrosis compatible with human NASH is desirable in models used for pharmacologic testing. When mice with specific genetic strains or mutations that cause over-eating consume a diet enriched with fat, modest amounts of cholesterol and/or simple sugars ("Western diet") they readily develop obesity with liver disease similar to human NASH, including significant fibrosis. Purely dietary models, such as high fat/high cholesterol, Western diet and choline-deficient amino acid defined are similarly promising. We share concern about using models without weight gain, adipose pathology or insulin resistance/hyperinsulinemia and with inadequate documentation of liver pathology. NASH-related fibrosis is a key endpoint in trials of possible therapies. When studied for this purpose, NASH models should be reproducible, show steatohepatitis (ideally with ballooning) and at least focal bridging fibrosis, while metabolic factors/disordered lipid partitioning should contribute to etiopathogenesis. Since murine models are increasingly used to explore pharmacologic therapies for NASH, we propose a minimum set of requirements that investigators, drug companies and journals should consider to optimise their translational value. This article is protected by copyright. All rights reserved.