The high failure rate of immunotherapies in multiple sclerosis (MS) clinical trials demonstrates problems in translating new treatment concepts from animal models to the patient. One main reason for this 'immunotherapy gap' is the usage of immunologically immature, microbiologically clean and genetically homogeneous rodent strains. Another reason is the artificial nature of the experimental autoimmune encephalomyelitis model, which favors CD4+ T cell driven autoimmune mechanisms, whereas CD8+ T cells are prevalent in MS lesions. In this paper, we discuss preclinical models in humanized rodents and non-human primates that are genetically closer to MS. We also discuss models that best reproduce specific aspects of MS pathology and how these can potentially improve preclinical selection of promising therapies from the discovery pipeline.