Animal models, by definition, are an approximation of reality, and their use in developing anti-cancer drugs is controversial. Positive retrospective clinical correlations have been identified with several animal models, in addition to limitations and a need for improvement. Model inadequacies include experimental designs that do not incorporate biological concepts, drug pharmacology, or toxicity. Ascites models have been found to identify drugs active against rapidly dividing tumors; however, neither ascitic nor transplantable subcutaneous tumors are predictive of activity for solid tumors. In contrast, primary human tumor xenografts have identified responsive tumor histiotypes if relevant pharmacodynamic and toxicological parameters were considered. Murine toxicology studies are also fundamental because they identify safe starting doses for phase I protocols. We recommend that future studies incorporate orthotopic and spontaneous metastasis models (syngeneic and xenogenic) because they incorporate microenvironmental interactions, in addition to confirmatory autochthonous models and/or genetically engineered models, for molecular therapeutics. Collectively, murine models are critical in drug development, but require a rational and hierarchical approach beginning with toxicology and pharmacology studies, progressing to human primary tumors to identify therapeutic targets and models of metastatic disease from resected orthotopic, primary tumors to compare drugs using rigorous, clinically relevant outcome parameters.