Malignant mesothelioma (MM), especially its more frequent form, malignant pleural mesothelioma (MPM), is a devastating thoracic cancer with limited therapeutic options. Recently, clinical trials that used immunotherapy strategies have yielded promising results, but the benefits are restricted to a limited number of patients. To develop new therapeutic strategies and define predictors of treatment response to existing therapy, better knowledge of the cellular and molecular mechanisms of MM tumors and sound preclinical models are needed. This review aims to provide an overview of our present knowledge and issues on both subjects. MM shows a complex pattern of molecular changes, including genetic, chromosomic, and epigenetic alterations. MM is also a heterogeneous cancer. The recently described molecular classifications for MPM could better consider inter-tumor heterogeneity, while histo-molecular gradients are an interesting way to consider both intra- and inter-tumor heterogeneities. Classical preclinical models are based on use of MM cell lines in culture or implanted in rodents, i.e., xenografts in immunosuppressed mice or isografts in syngeneic rodents to assess the anti-tumor immune response. Recent developments are tumoroids, patient-derived xenografts (PDX), xenografts in humanized mice, and genetically modified mice (GEM) that carry mutations identified in human MM tumor cells. Multicellular tumor spheroids are an interesting in vitro model to reduce animal experimentation; they are more accessible than tumoroids. They could be relevant, especially if they are co-cultured with stromal and immune cells to partially reproduce the human microenvironment. Even if preclinical models have allowed for major advances, they show several limitations: (i) the anatomical and biological tumor microenvironments are incompletely reproduced; (ii) the intra-tumor heterogeneity and immunological contexts are not fully reconstructed; and (iii) the inter-tumor heterogeneity is insufficiently considered. Given that these limitations vary according to the models, preclinical models must be carefully selected depending on the objectives of the experiments. New approaches, such as organ-on-a-chip technologies or in silico biological systems, should be explored in MM research. More pertinent cell models, based on our knowledge on mesothelial carcinogenesis and considering MM heterogeneity, need to be developed. These endeavors are mandatory to implement efficient precision medicine for MM.