Osteosarcoma (OS) is a highly aggressive tumor characterized by malignant cells producing pathologic bone; the disease presents a natural tendency to metastasize. Genetic studies indicate that the OS genome is extremely complex, presenting signs of macro-evolution, and linear and branched patterns of clonal development. However, those studies were based on the phylogenetic reconstruction of next-generation sequencing (NGS) data, which present important limitations. Thus, testing clonal evolution in experimental models could be useful for validating this hypothesis. In the present study, lentiviral LeGO-vectors were employed to generate colorimetric red, green, blue (RGB)-marking in murine, canine, and human OS. With this strategy, we studied tumor heterogeneity and the clonal dynamics occurring in vivo in immunodeficient NOD.Cg-Prkdcscid-Il2rgtm1Wjl/SzJ (NSG) mice. Based on colorimetric label, tumor clonal composition was analyzed by confocal microscopy, flow cytometry, and different types of supervised and unsupervised clonal analyses. With this approach, we observed a consistent reduction in the clonal composition of RGB-marked tumors and identified evident clonal selection at the first passage in immunodeficient mice. Furthermore, we also demonstrated that OS could follow a neutral model of growth, where the disease is defined by the coexistence of different tumor sub-clones. Our study demonstrates the importance of rigorous testing of the selective forces in commonly used experimental models.