Grasslands in Inner Mongolia have been confronted with unprecedented degradation in recent years. Research on ecosystem stability is important to inform evaluation of the health of degraded grassland ecosystems. We examined synthetic stability, which was defined by the relationships between multiple components of stability, known as multidimensional stability, in grasslands at four stages of degradation (undegraded, slightly degraded, moderately degraded, and intensely degraded) in the Xilin River Basin, Inner Mongolia, China. We analyzed the connections between multidimensional stability and its relationship with four stability components, including community resistance and recovery measured on the basis of plant functional traits, and community functional (aboveground net primary productivity) and structural (Jaccard dissimilarity) variability, calculated on the basis of ten plots from different spatial distributions in a study site. Our results showed that (i) 9 of 17 traits displayed a significant trend along the grazing intensity gradient, indicating a clear turnover of species within communities in response to the grazing intensity gradient; (ii) moderately degraded (C) grasslands showed higher recovery, resistance, and synthetic stability than undegraded (A) and slightly degraded (B) grassland communities overall (recovery: p = 0.026, p = 0.032, for pairs of samples from A and C, and B and C, respectively; resistance: p = 0.024, for a pair of samples from A and C), which conformed with the intermediate disturbance hypothesis and positive diversity-stability relationship; and (iii) the multidimensionality of stability varied between different stages of degradation and were strongly dependent upon the correlations between stability components. Our study is expected to enrich the theory of stability maintenance in grassland ecosystems and provide guidance for grassland restoration and biodiversity conservation.