Grazing is very common in the grassland ecosystem, and it has a significant impact on the C stocks and cycle. One of the most important drivers of soil C stocks is functional diversity. However, limited studies have attempted to explore the effects of functional diversity on soil C stocks associated with grazing disturbance. This study was carried out in Hulunbeier grassland, Inner Mongolia, and four grazing intensities (no grazing (NG), light grazing (LG), moderate grazing (MG), and heavy grazing (HG)) were identified. The plant functional traits and important soil properties under different grazing intensities were measured. Functional identity and diversity were calculated based on the measured functional traits. The impacts of functional identity and diversity on soil organic carbon stocks (SOCstocks) were analyzed using a multi-model inference (MMI) approach. Our study showed that the functional diversity effect on soil C stocks varies depending on grazing intensity. We identified that functional richness has a significant impact on SOCstocks in NG. The community weighted mean of leaf area became the best predictor of SOCstocks in LG. As grazing intensified, functional divergence best explained SOCstocks in moderate and heavy grazing sites, and their relationship was positive. The major outcomes of this research could shed light on the mechanics of soil carbon storage.