Plant species diversity has long been a focal point in ecological studies. In order to study the changes in species diversity at different spatial scales (alpha, beta, and gamma diversities) in the restoration process of grassland vegetation in fragile desert steps, this study took desert steppe of Inner Mongolia as the research object and employed a two-factor experimental design that combined exclosure years (the years when an area was isolated to prevent grazing and other disturbances) with years of monitoring (the years when data were collected). It analyzed the plant groups (dominant species, common species, and rare species) and species diversity, and obtained the preliminary conclusions as follows: The optimal exclosure duration for promoting species diversity balance in desert steppe management is between 16 and 18 years. Short-term exclosure enhances species diversity by promoting recovery in overgrazed systems, while long-term exclosure may reduce diversity due to dominant species proliferation and inhibited regeneration. Increasing the duration of exclosure (the period from the initial exclosure year to the year of monitoring) can improve plant species diversity. Exclosure years and years of monitoring exhibited a significantly positive influence on alpha, beta, and gamma diversities, with a negative interaction effect between exclosure years and years of monitoring. In addition, plant groups played a significant role in diversity at different spatial scales. Contribution to alpha diversity ranked as follows: rare species > common species > dominant species; contribution to beta diversity ranked as rare species > dominant species > common species; contribution to gamma diversity ranked as common species > dominant species > rare species. Rare species played a crucial role in maintaining diversity stability within the community and diminishing gradient differences, and common species were instrumental in upholding landscape features.