The relationship between sampling area of plant community and number of species contained in it has an importance in vegetation ecology. Using a nested quadrat survey, the relationship between plant species and sampling area was discussed under different grazing pressure in the steppe and the determination of minimum sampling area was calculated herein. Ten curve models were selected to fit the species-area relation of plant community in a typical steppe of northern China. The residual standard deviation, correlation index, average value of absolute deviation, and average value of absolute relative deviation were used as the evaluation indexes for model fitting. The results show that the curve equations S = B + CAZ and S = C(1-e−BA)Z have the best simulation effect. According to the simulation results, the minimum sampling areas are determined as 1.16 m2 (T0), 1.23 m2 (T7), 1.28 m2 (T10), and 1.10 m2 (T14), presenting the minimum sampling area is feasible when the second derivative is zero. Generally, sheep grazing reduced the species richness of steppe, though some of the plans flourished while some others disappeared or dismissed, but increased spatial heterogeneity, especially on a smaller scale. However, there are also variances in different area scales and grazing intensities. When the sampling area was small (< 2 m2), heavy grazing T14 increased the number of plant species per unit area, yet light grazing T8 reduced the number of plant species per unit area of steppe. When the sampling area was large (> 16 m2), grazing reduced the number of plant species per unit area in the natural steppe. Anyhow, the effect of grazing intensity on plant species fluctuated greatly when the sampling area was medium (2 m2 ∼ 16 m2). In general, the similarity of plant communities had an increasing trend with accumulated sampling area, though it varied at different grazing intensities.