Immunotherapy is revolutionizing cancer treatment; however, complete responses are achieved in only a small fraction of patients and tumor types. Thus, there is an urgent need for predictive preclinical models to drive rational immunotherapeutic drug development, treatment combinations, and to minimize failures in clinical trials. Humanized mouse models (HIS) have been developed to study and modulate the interactions between immune components and tumors of human origin. In this review, we discuss recent advances in the ‘humanization’ of mouse models to improve the quality of human immune cell reconstitution. We also highlight new insights into the basic mechanisms, and provide a preclinical evaluation of onco-immunotherapies, as well as the limitations thereof, which constitute drivers for the improvement of the models to increase their translational power. Cancer immunotherapy has given impressive clinical responses but only in a minority of patients and only in certain cancer indications. Thus, predictive preclinical models are needed to drive rational immunotherapeutic drug development and minimize failures in clinical trials. HIS, comprising immunodeficient host mice and human immune and tumor cells, recapitulate the interactions between immune components and tumors of human origin. Limitations of HIS models include the development of xenograft-versus-host disease and incomplete reconstitution of certain human immune subpopulations. HIS models have allowed the evaluation of onco-immunotherapies, including cellular and antibody-based immunotherapies. Understanding the caveats of HIS mice and the increasing genetic optimizations are leading to improved models with heightened translational power.