Disappointing results from most late-stage clinical trials of cancer therapeutics indicate a need for improved and more-predictive animal tumor models. This insufficiency of models, combined with the advent of a class of drugs that target the tumor microenvironment rather than the tumor cell, presents new challenges for designing and interpreting preclinical efficacy studies. A comparison of the clinical efficacy of anti-angiogenic drugs with their corresponding preclinical studies over the past two decades offers many lessons that can inform and improve the design of experiments in existing mouse models. In addition, technological and logistical advances in mouse models of human cancer over the past five years have the potential to increase the clinical translatability of animal studies.