Despite major advances in recent years, graft-versus-host disease (GVHD) remains a major life-threatening complication of allogeneic hematopoietic cell transplantation (allo-HCT). To improve our therapeutic armory against GVHD, preclinical evidence is most frequently generated in mouse and large animal models of GVHD. However, because every model has shortcomings, it is important to understand how predictive the different models are and why certain findings in these models could not be translated into the clinic. Weaknesses of the animal GVHD models include the irradiation only-based conditioning regimen, the homogenous donor/recipient genetics in mice, canine or non-human primates (NHP), anatomic site of T cells used for transfer in mice, the homogenous microbial environment in mice housed under specific pathogen-free conditions, and the lack of pharmacologic GVHD prevention in control groups. Despite these major differences toward clinical allo-HCT, findings generated in animal models of GVHD have led to the current gold standards for GVHD prophylaxis and therapy. The homogenous nature of the preclinical models allows for reproducibility, which is key for the characterization of the role of a new cytokine, chemokine, transcription factor, microRNA, kinase, or immune cell population in the context of GVHD. Therefore, when carefully balancing reasons to apply small and large animal models, it becomes evident that they are valuable tools to generate preclinical hypotheses, which then have to be rigorously evaluated in the clinical setting. In this study, we discuss several clinical approaches that were motivated by preclinical evidence, novel NHP models and their advantages, and highlight the recent advances in understanding the pathophysiology of GVHD.