Knowledge about the genetic diversity of a specie is essential for plant breeding. The aim of this study was to estimate the genetic diversity among Brazilian grape cultivars from the Grapevine Active Germplasm Bank of Embrapa Semi & aacute;rido focusing on the quality attributes and bioactive compounds of the fruit using self-organizing maps. Thirty-six grape cultivars were evaluated over two consecutive crop seasons considering the following variables: berry color (lightness, chroma, and hue angle), berry firmness, soluble solids, titratable acidity, soluble solids/titratable acidity ratio, total extractable polyphenols, yellow flavonoids, and total anthocyanins. The variance components were estimated using restricted maximum likelihood/best linear unbiased prediction (REML/BLUP). Cluster analysis using self-organizing maps was performed based on Euclidean distance. Estimates of individual repeatability (r) ranged from 0.40 to 0.93, and were considered high for most of the variables evaluated. Mean selective accuracy was high, ranging from 0.76 to 0.97. Cluster analysis showed the formation of five groups. The proximity among Groups 1, 2, and 3 is explained by their sharing the importance of the lightness and chroma variables, and proximity between Groups 4 and 5 by hue angle, berry firmness, soluble-solids content, anthocyanins, and soluble solids/titratable acidity ratio. REML/BLUP made it possible to obtain accurate estimates of genetic diversity among the grape cultivars evaluated. The genetic variability among the Brazilian grape cultivars will allow identification of genotypes with high content quality atributes for use in future research efforts, especially recommendation for grape growers and development of new grape cultivars for the Brazilian semi-arid region.