Spanish wheat (Triticum spp.) landraces have a considerable polymorphism, containing many unique alleles, relative to other collections. The existence of a core collection is a favored approach for breeders to efficiently explore novel variation and enhance the use of germplasm. In this study, the Spanish durum wheat (Triticum turgidum L.) core collection (CC) was created using a population structure–based method, grouping accessions by subspecies and allocating the number of genotypes among populations according to the diversity of simple sequence repeat (SSR) markers. The CC of 94 genotypes was established, which accounted for 17% of the accessions in the entire collection. An alternative core collection (CH), with the same number of genotypes per subspecies and maximizing the coverage of SSR alleles, was assembled with the Core Hunter software. The quality of both core collections was compared with a random core collection and evaluated using geographic, agromorphological, and molecular marker data not previously used in the selection of genotypes. Both core collections had a high genetic representativeness, which validated their sampling strategies. Geographic and agromorphological variation, phenotypic correlations, and gliadin alleles of the original collection were more accurately depicted by the CC. Diversity arrays technology (DArT) markers revealed that the CC included genotypes less similar than the CH. Although more SSR alleles were retained by the CH (94%) than by the CC (91%), the results showed that the CC was better than CH for breeding purposes.