Plant diversity is a crucial component of biodiversity, and it has attracted growing attention in recent years. For desert steppes, however, how grazing and the sample area used in these studies affect species diversity, and through what mechanisms, is still unclear. Stipa breviflora grasslands typically represent desert steppes, which in turn occupy a special place among the grassland ecosystems. In this study, a long-term (11 years) grazing experiment with four levels of grazer stocking rate in the S. breviflora desert steppe in Siziwang Banner of Inner Mongolia was used. The study aims to provide a scientific basis for grassland biodiversity protection and sustainable development, and therefore explored the influence of different grazer stocking rates on plant diversity (using four alpha diversity indexes:the Margalef, Shannon-Wiener, Simpson, and Pielou index). Grassland alpha diversity and species composition change was assessed for different stocking rates, and the sample area sizes (from 0.01 m² to 650 m²) were analyzed to determine the optimal sampling area for this desert grassland.The results showed that:(1) The number of species and alpha diversity indices declined with increasing stocking rate. (2) Compared with the relative density of the dominant species (e.g. S. breviflora and Cleistogenes songorica), grazing reduced the relative density of non-dominant species. Dominant species have a strong adaptability and endurance, whereas non-dominant species have a sensitive response to grazing. Moreover, a decrease in non-dominant species resulted in a lower alpha diversity index. This suggests that non-dominant species are important indicators that reflect how plant species respond to grazer stocking rate changes, and underline the variety in adaptive strategies for resources changes. (3) The best sampling area size to investigate the species number and alpha diversity was 40 m². (4) The species-area relations and alpha diversity index-area relations both conformed to a logarithmic growth model. (5) Sampling area influenced the analyses of the different alpha diversity indices. With increasing scale area, the differences among the alpha diversity indices increased gradually with stocking rate gradients. Sampling areas between 0.16 m² to 0.64 m² (small scale) detected significant different levels of alpha diversity indices between zero and low, and moderate and high stocking rate levels. However, sampling areas larger than 160 m² (large scales) could detect significant differences in alpha diversity indices among high and all stocking rate levels over the sampling area.