Animal models of human cancer have evolved in attempts to capture the complexity of the human disease. They encompass two broad types of model, namely those in which the tumor arises in situ and those in which cancer cells or tissue are transplanted. Currently human tumor xenografts are the most widely used model to help predict antitumor efficacy in a preclinical setting and xenograft results for certain disease types such as ovarian cancer and non-small cell lung cancer correlate well with clinical activity if the models are used under appropriate conditions. The genetically engineered mouse (GEM) models allow study of the effects of targeted inhibitors against defined molecular targets. These are becoming increasingly sophisticated to recapitulate the progression of tissue-specific molecular changes found within individual cancers. Non-germline GEM models possess the ability to study the impact of specific cancer genes without some of the limitations inherent in traditional GEM models. While rodents, particularly mice, have been the animal host most frequently used, there is increasing recognition of the value of using larger species especially dogs in the veterinary oncology setting. These have successfully modelled aspects of selected human cancers such as osteosarcoma, GIST and prostate cancer. Individually, these models have relative strengths and weaknesses in mimicking the human disease and appropriately reflecting its cellular and molecular pathology. This review will seek to address where these models can be best used to help predict response to therapeutics.