<p>The establishment of core germplasm collections is a pivotal strategy for genetic improvement of cultivars. In China, where sorghum serves as a critical crop predominantly for liquor production, the utilization of its genetic resources is paramount to the distilling industry. However, no core collection has been developed specifically for sorghum resources in the Chishui River Basin, the epicenter of China’s premium Jiang-flavor Baijiu production. This study included 1,309 local sorghum accessions from the region, employing data from 18 agronomic traits. Through a comparative analysis of various sampling methods, proportions and clustering techniques, we determined that the “multi-clustering deviation sampling method” combined with a 5% sampling ratio, Euclidean distance, and the farthest neighbor algorithm was optimal for constructing the core sorghum germplasm of the Chishui River Basin. The resultant core germplasm, comprising 75 accessions or 5.73% of the total germplasm, showed a mean difference percentage, variance difference percentage, range compliance rate and coefficient of variation change rate of 0%, 82.15%, 96.54% and 118.23%, respectively, relative to the entire sample set. The representativeness of these 75 core accessions was validated through <i>t</i>-tests comparing means, extremes, and standard deviations across different traits, as well as through genetic diversity index <i>t</i>-value tests and principal component analysis. This core collection effectively captures the genetic characteristics and population structure of the full germplasm pool, laying a solid foundation for sorghum cultivar improvement, gene discovery and germplasm innovation in the Chishui River Basin.</p>