Recent studies cast doubt on the value of traditionally used models as tools for testing therapies for human cancer. Although the standard practice of xenografting tumors into immunocompromised mice generates reproducible tumors, drug testing in these models has low predictive power when compared to the clinical responses in Phase II trials. The use of tumor-bearing genetically engineered mouse models holds promise for improving preclinical testing. These models recapitulate specific molecular pathways in tumor initiation or progression and provide a biological system in which to study the disease process for assessing efficacy of new therapies and proof-of-principle for testing molecularly targeted drugs. In this review, we discuss the advantages and limitations of genetically engineered mice and plausible solutions for adapting these valuable tumors for wider use in preclinical testing by transplantation into naïve recipients. We also provide examples of comparative molecular analysis of mammary tumors from MMTV-Polyoma Middle-T antigen and MMTV-wnt1 models as tools for finding clinical correlates, validating existing models and guiding the development of new genetically engineered mouse models for cancer.