Primary lymphoma of the central nervous system (CNS, PCNSL) is a specific diffuse large B cell lymphoma entity arising in and confined to the CNS. Despite extensive research since many decades, the pathogenetic mechanisms underlying the remarkable tropism of this peculiar malignant hematopoietic tumor remain still to be elucidated. In the present review, we summarize the present knowledge on the genotypic and phenotypic characteristics of the tumor cells of PCNSL, give an overview over deregulated molecular pathways in PCNSL and present recent progress in the field of preclinical modeling of PCNSL in mice. With regard to the phenotype, PCNSL cells resemble late germinal center exit IgM+IgD+ B cells with blocked terminal B cell differentiation. They show continued BCL6 activity in line with ongoing activity of the germinal center program. This together with the pathways deregulated by genetic alterations may foster B cell activation and brisk proliferation, which correlated with the simultaneous MYC and BCL2 overexpression characteristic for PCNSL. On the genetic level, PCNSL are characterized by ongoing aberrant somatic hypermutation that, besides the IG locus, targets the PAX5, TTF, MYC, and PIM1 genes. Moreover, PCNSL cells show impaired IG class switch due to sμ region deletions, and PRDM1 mutations. Several important pathways, i.e., the B cell receptor (BCR), the toll-like receptor, and the nuclear factor-κB pathway, are activated frequently due to genetic changes affecting genes like CD79B, SHIP, CBL, BLNK, CARD11, MALT1, BCL2, and MYD88. These changes likely foster tumor cell survival. Nevertheless, many of these features are also present in subsets of systemic DLBLC and might not be the only reasons for the peculiar tropism of PCNSL. Here, preclinical animal models that closely mimic the clinical course and neuropathology of human PCNSL may provide further insight and we discuss recent advances in this field. Such models enable us to understand the pathogenetic interaction between the malignant B cells, resident cell populations of the CNS, and the associated inflammatory infiltrate. Indeed, the immunophenotype of the CNS as well as tumor cell characteristics and intracerebral interactions may create a micromilieu particularly conducive to PCNSL that may foster aggressiveness of tumor cells and accelerate the fatal course of disease. Suitable animal models may also serve as a well-defined preclinical system and may provide a useful tool for developing new specific therapeutic strategies.