Assessment of spring snow cover fraction (SCF) can be valuable for understanding the efficacy of certain Earth system models (ESMs) in simulating the energy exchange between land and atmosphere system, global hydrological cycle and future climate impacts. In this work, we studied the model performance of 23 and 20 ESMs participating CMIP5 and CMIP6 by comparing satellite data from National Oceanic and Atmospheric Administration and National Climatic Data Center (NOAA/NCDC) over the Northern Hemisphere (NH) and its 13 sub-regions and further evaluating potential model improvement from CMIP5 to CMIP6. We found that the mean annual spring SCF that was simulated by CMIP5 and CMIP6 models was underestimated in most of the NH and overestimated on the Tibetan Plateau and in eastern Asia, and most of the model simulations showed a stronger reduction trend as well as an underestimated monthly climatological SCF. However, compared with those in CMIP5, most of the model simulations in CMIP6 had an improved ability to simulate spring SCF in terms of the annual mean, long-term trend and intra-annual variability. We also confirmed that the multi-model ensemble mean (MME) is a better way to represent the three aspects of spring SCF than most individual model simulations. The spring SCF values predicted by the CMIP5 and CMIP6 MMEs over the NH and its 13 sub-regions under different scenarios showed decreasing trends. The decreasing spring SCF trends differed under different scenarios, and the SCF under high emissions scenarios (RCP 8.5 and SSP5-8.5) continued to decrease.