Technical Abstract: A rice mini-core collection consisting of 217 accessions has been developed to represent the USDA core and whole collections that include 1,794 and 18,709 accessions, respectively. To improve the efficiency of mining valuable genes and broadening the genetic diversity in breeding, genetic structure and diversity were analyzed using both genotypic (128 molecular markers) and phenotypic (14 numerical traits) data. This mini-core had 13.5 alleles per locus, which is the most among the reported germplasm collections of rice. Similarly, polymorphic information content (PIC) value was 0.71 in the mini-core which is the highest with one exception. The high genetic diversity in the mini-core suggests there is a good possibility of mining genes of interest and selecting parents which will improve food production and quality. A model-based clustering analysis resulted in lowland rice including three groups, aus (39 accessions), indica (71) and their admixtures (5), upland rice including temperate japonica (32), tropical japonica (40), aromatic (6) and their admixtures (12) and wild rice (12) including glaberrima and four other species of Oryza. Group differentiation was analyzed using both genotypic distance Fst from 128 molecular markers and phenotypic (Mahalanobis) distance D2 from 14 traits. Both dendrograms built by Fst and D2 reached similar-differentiative relationship among these genetic groups. The correlation coefficient of genotypic distance with phenotypic distance among the groups was 0.85, which is much higher than it in cassava (Manihot esculenta Crantz), the only another crop in which differentiation has been studied by both genotypic and phenotypic characterization. Genetic differentiation could be helpful for the association mapping of genes of interest. Precautions for genetic differentiation using phenotypic characteristics were discussed.