The increase in energy consumption in Brazil and in the world generates an increasing need to seek renewable and non-polluting energy, such as biofuels. Safflower (Carthamus tinctorius L.) is a plant with a great oil production capacity and potential for biodiesel production that presents high productivity and easy edaphoclimatic adaptation. It can be an economic culture option in crop rotation practiced by farmers, however it is still necessary to obtain more technical information about its cultivation and about the adapted and improved cultivars. Thus, the estimate of genetic divergence from morphological characters using multivariate techniques has become a common alternative among breeders. In this sense, the objective of this research was to carry out an analysis of the genetic divergence among the 49 safflower genotypes from the germplasm bank of the Instituto Mato Grossense do Algodão (IMA-MT), based on 13 morphological descriptors recommended by the International Board for Plant Genetic Resources - IBPGR (1983) and Ministério da Agricultura, Pecuária e Abastecimento - MAPA (2013). The divergence analysis was performed by means of the dissimilarity matrix based on multi-categorical variables and to identify the most similar genotypes, Tocher's optimization grouping methods and the hierarchical average grouping method between groups were used. The estimates of the dissimilarity coefficients ranged from 0.00 to 0.46, indicating the presence of genetic diversity among the evaluated genotypes. The highest dissimilarity value was between genotypes 29 and 9, 42 and 1, 42 and 32, 47 and 9, which are the most genetically divergent and thus promising materials for future genetic crosses. The UPGMA dendrogram and tocher grouping were partially consistent and effective in grouping safflower genotypes.