There is tremendous need for improved prostate cancer (PCa) models. The mouse prostate does not spontaneously form tumors and is anatomically and developmentally different from the human prostate. Engineered mouse models lack the heterogeneity of human cancer and rarely establish metastatic growth. Human xenografts represent an alternative but rely on an immunocompromised host. Accordingly, we generated PCa murine xenograft models with an intact human immune system (huNOG and huNOG-EXL mice) to test whether humanizing tumor-immune interactions would improve modeling of metastatic PCa and the impact of hormonal and immunotherapies. These mice maintain multiple human cell lineages, including functional human T-cells and myeloid cells. In 22Rv1 xenografts, subcutaneous tumor size was not significantly altered across conditions; however, metastasis to secondary sites differed in castrate huNOG vs background-matched immunocompromised mice treated with enzalutamide (enza). VCaP xenograft tumors showed decreases in growth with enza and anti-Programed-Death-1 treatments in huNOG mice, and no effect was seen with treatment in NOG mice. Enza responses in huNOG and NOG mice were distinct and associated with increased T-cells within tumors of enza treated huNOG mice, and increased T-cell activation. In huNOG-EXL mice, which support human myeloid development, there was a strong population of immunosuppressive regulatory T-cells and Myeloid-Derived-Suppressor-Cells (MDSCs), and enza treatment showed no difference in metastasis. Results illustrate, to our knowledge, the first model of human PCa that metastasizes to clinically relevant locations, has an intact human immune system, responds appropriately to standard-of-care hormonal therapies, and can model both an immunosuppressive and checkpoint-inhibition responsive immune microenvironment.