The objective of this work was to quantify the genetic variability of 213 conilon coffee accessions of the Robusta Tropical cultivar, based on chemical characteristics related to the quality of green coffee beans, as well as to identify promising accessions for the breeding program of irrigated conilon coffee in the Brazilian Cerrado. The chemical characteristics evaluated were: protein and caffeine contents, total soluble solids and total lipids, ether extract, pH, and total titratable acidity. The data were subjected to the principal component analysis and cluster analysis based on the similarities observed within the first two principal components using the minimum variance method (Wards) and, as a measure of similarity, the Euclidean distance. The three main components explained 72.64% of the total variation of the data. All characteristics, except pH, were correlated with the first three components. It was possible to separate the genotypes in three clusters, according to the similarities observed in the behavior of the variables. The evaluated accessions present genetic variability regarding the assessed quality characteristics of green coffee beans, and CPAC 160 and CPAC 32 are the most promising for the breeding program of conilon coffee for cultivation under irrigation in the Cerrado.