In the context of “Carbon Emissions Peak” and “Carbon Neutrality”, grazing exclusion (GE) has been applied widely to rehabilitate degraded grasslands and increase carbon sequestration. However, on the Qingzang Plateau (QP), the impacts of GE on carbon dynamics of alpine grasslands are poorly understood, particularly at regional scale. Here, we evaluated the responses of carbon sequestration to GE in different alpine grasslands across QP to fill this knowledge gap by using meta-analysis. Overall, the effects of GE on ecosystem carbon fractions were dependent on GE duration, grassland types and climate factors. GE had more obviously positive effects on carbon stock across alpine meadow than alpine steppe. Longer duration of GE was more effective for ecosystem carbon sequestration in alpine steppe. Annual mean precipitation (AMP) and temperature (AMT) began to dominate ecosystem carbon sequestration after three years of GE duration across alpine meadow; and AMP was important climate factor limiting ecosystem carbon sequestration in the alpine steppe. In terms of plant carbon fraction, GE generated continuous positive effects on aboveground biomass (AGB) with increased GE duration in alpine meadow, while the continuous beneficial effects for AGB of alpine steppe disappeared at the 8th year of GE duration. And no positive effects were found on belowground biomass at the 11th year in both alpine meadow and alpine steppe. For soil organic carbon and microbial biomass carbon, there were no obvious trends in response to GE duration. In general, we highlighted that the responses of different carbon fractions (plant -soil - microbe) to GE were nonuniform at spatial and temporal scales, thereby we should adopt different carbon management practices for sustainable development of different grasslands.