Simulation of the spring phenology of grasslands is crucial for understanding how ecosystem respond to climate change and for precisely modeling the carbon, water, and energy balance in terrestrial ecosystems. In this study, the performance of 15 phenology models was compared in terms of their ability to simulate the spring phenology of three grassland types in Central Asia from 2000 to 2019, using meteorological data and remote sensing inversion phenology data. The start of the grassland season (SOS) was simulated, and the root mean square errors (RMSEs) were evaluated. The phenology models produced a median SOS RMSE of 6.7 +/- 2.5 days. Among the 15 models, the Temperature-precipitation model (TP), Growing degree day model (GDD), and Temperatureprecipitation sequential model (TPS) demonstrated superior performance for temperate grassland, desert grassland, and mountain meadow, respectively. Their respective median RMSEs were 4.20, 5.37, and 5.28 days. Furthermore, the study also projected the future grassland SOS from 2020 to 2100 under seven radiative forcing scenarios using these three models. The simulation results indicated that the SOS for all three grassland types would advance in the future, especially under higher radiative forcing scenarios. The study provides a scientific basis for the sustainable development of grassland ecosystems in Central Asia.