Abstract: The objective of this research was to study the stability of yield and test-site representativeness of proso-millet cultivars in China. Six cultivars were tested in nine sites in 10 m2 experimental plot under randomized completed block design and three replicates. The combination of double mark figure and special interaction Dg(e) of AMMI (additive main effects and multiplicative interaction) model was used to identify yield stability and test-site representativeness of the proso-millet cultivars. Analyzed data were from the national regional trials of proso-millet for 2006—2008. The results showed that the proportions of the sum of squares of genotype (G), environment (E) and G×E interactive effects on total sum of squares were 2.76%, 6.26% and 52.85%, respectively. There were significant effects in G action, E action and G×E interaction. IPCA1, IPCA2, IPCA3 accounted for up to 92.58% of the overall interaction effects. There were significant differences in stabilities of the six cultivars and representativeness of the nine test-sites. Among the six cultivars, “Yumi No.3” (CK) and “Yi8414-1-2-1” were high yield and better yield stability cultivars. “Gan9109-6-1-1-2” and “Gu01-391” exhibited less stability and high yield production. “Gan9133-1-3-4-1” showed to be a high-stability less-yield cultivar and “Gu02-25” was a low-yield unstable cultivar. For the nine test-sites, Fugu (Shaanxi Province), Tongxin (Ningxia Hui Autonomous Region), Yulin (Shaanxi Province) and Chifeng (Inner Mongolia Autonomous Region) had a better discrimination than Guyuan and Yanchi (Ningxia Hui Autonomous Region), Huining (Gansu Province), Wuzhai (Shanxi Province) and Eerduosi (Inner Mongolia Autonomous Region).