Genotype stability of agronomic traits is a direct measure of the effect of genotype × environment interactions (GEI). The additive main effects and multiplicative interaction model (AMMI) combine analysis of variance and principal component analyses that has been proved to be an effective tool in understanding complex GEI and evaluating genotypes stability and environmental discrimination to genotype diversity. The grain yield data of 10 spring wheat genotypes grown in a regional trial of China at 13 sites of the rainfed areas were used for AMMI analysis. The average grain yield across the 13 sites ranged from 396.6 kg•hm^－２ to 4050.2 kg•hm^－２, of which the grain yields of newly released, modern genotypes ranged from 1318.6 kg•hm^－２ to 2315.6 kg•hm^－２. Highly significant variations were observed among genotypes, environments and GEI, with the three variances accounting for 6.2%, 70.3% and 23.5% of total variance, respectively. The environment and the GEI had greater effects on grain yield than genotypes. The three first principal component axes (IPCA) with significant interaction represented 90.8% of total GEI variance. Using the three first IPCA, genotype stability parameter (Di) and site discrimination parameter (Dj) were calculated, where the difference of Di among the genotypes was up to 3 times, while that of Dj among the sites was up to 19 times; Genotypes Dingxi 35, Xihan No.1 and Dingfeng 889 had highest yields and best yield stability across different environments. The genotype Mengmai 35 was the worst in both yield and stability. Some genotypes performed best at some specific sites and can be recommended for those specific environments.