The grassland ecosystem of the Mongolian Plateau is a significant carbon sink and highly sensitive to environmental changes. Monitoring gross primary productivity (GPP), net primary productivity (NPP), and carbon use efficiency (CUE), along with their responses to climate changes, is essential for accurate carbon budgeting and forecasting. Our study, conducted from 2000 to 2018, we observed fluctuating yet upward trends in NPP, GPP and CUE across the grasslands, indicating an increased potential for carbon sequestration, while the spatial distribution of NPP and GPP showed a notable increasing trend from the west to the east. By employing an interpretable machine learning approach, we correlated GPP, NPP and CUE with key climatic variables-air temperature, precipitation, soil moisture (SM), and solar radiation (SR)-and quantified their individual impacts. Our findings reveal that GPP, NPP, and CUE in grasslands are positively influenced by precipitation and SM, but negatively by air temperature and SR. Precipitation contributes 50.14% to GPP and 49.63% to NPP, while SM contributes 19.3% to GPP and 18.18% to NPP. In the case of CUE, SM is the predominant factor, affecting 28.3 % of the area, followed by air temperature, which affects 27.94 % of the area. In contrast, precipitation and SM have a comparatively lesser impact on CUE, affecting 24.88 % and 18.84% of the area, respectively. The findings of this research are extremely important for gaining a thorough understanding of the carbon cycle response mechanism of grassland vegetation in dry and semi-arid regions in the context of climate change.