Based on 324 germplasm resources from the Taigu Grape Nursery， a national grape germplasm repository， multivariate statistical methods were used to systematically evaluate the genetic diversity of 18 fruit traits. Phenotypic variation analysis revealed the coefficients of variation （CV） ranging from 10.54% to 63.58% in 18 traits including berry weight， flesh texture， and hardness. The Shannon-Wiener indices （H′） ranged from 0.69 to 2.07， and among these， bunch length （H′=2.05）， bunch width （H′=2.07）， berry height （H′=2.05）， and berry width （H′=2.06） exhibited high genetic diversity， indicating significant phenotypic differences and diverse types within the germplasm collection. Correlation analysis revealed that berry weight was significantly and positively correlated with traits such as berry height， berry width， and hardness， thus serving as a core selection index for large-berry breeding. Cluster analysis using Euclidean distance （threshold=13.605） classified the 324 accessions into four distinct groups： Group I （high-quality table grapes）， Group II （dual-purpose wine and table）， Group III （wine-making type with storage resistance）， and Group IV （large-berry breeding type）. Principal component analysis （PCA） extracted six principal components （PCs）， collectively explaining 80.083% of the total variance. PC1 represented the morphology-hardness dimension， PC2 the texture-bunch architecture dimension， and PC3 the seed development dimension. Based on comprehensive PCA scores， we identified the top ten excellent germplasms， including table grape cultivars such as ‘Ribier’ and ‘Pink Grape’. Additionally， 20 specific germplasms meeting defined criteria （berry weight>9 g， seed number≤1， hardness>1900 N， springiness>0.8） were identified， providing elite parental lines and key genetic resources for targeted grape breeding programs.