Cancer is a multigenic pathology whose dynamic, evolving and heterogeneous nature has led to the development of a variety of models to study it. Traditionally, two-dimensional (2D) cell culture has been used in vitro. However, an accumulated body of data demonstrates the relevance of three-dimensional (3D) cultures for modeling and studying the mechanisms involved in the initiation, progression, and resistance to treatments of tumors, particularly solid tumors. By better reproducing the spatial architecture, cell-cell and cell-matrix interactions, and phenotypic heterogeneity of tumors, these 3D models help bridge the gap between in vitro studies and in vivo conditions. This review describes the main 3D culture models - spheroids, organoids, tumoroids, tumor explants and cancers-on-a-chip - highlighting their principles, advantages and limitations. The techniques used to generate these 3D culture models, such as liquid-based or scaffold-based 3D culture, microfluidics, and 3D bioprinting, are also presented. Finally, examples of applications of these models are discussed, including the study of vascularization and its role in immune system recruitment, the modeling of the metastatic process, the screening of drug candidates and the elucidation of treatment resistance. The potential of 3D models is also addressed in the context of personalized medicine.