Abstract: In this research, output from eight Chinese Earth climate system models from the International Coupled Model Intercomparison Project Phase 6 (CMIP6) were selected. The spatial distribution and long-term trend of the multi-year mean Arctic sea ice thickness between 1980 to 2014 from the models were evaluated by comparing them with the sea ice thickness data assimilation product from the Pan-Arctic Ice-Ocean and Assimilating System (PIOMAS). The capability of each model was quantified using the Taylor Score. The results show that there are differences between output from all models and PIOMAS data in both March and September, and the deviations are mainly located in the north of Greenland, the Barents Sea, and around the Bering Strait. Models underestimate March and September ice thickness of the Arctic central region in, and generally overestimated the thickness of the Arctic sea ice margin region in March. In terms of spatial distribution of ice thickness, the difference between PIOMAS data and FIO-ESM-2-0 output is the smallest among the eight models and the difference between PIOMAS and FGOALS-f3-L is the second smallest. The long-term trend of sea ice thickness is negative, with sea ice thickness decreasing faster, in September than in March. In terms of the spatial distribution of the long-term sea ice thickness trend, the difference between PIOMAS data and model output is the smallest for NESM3 among the eight models. The radiation components from the eight CMIP6 models were compared with three reanalysis products and one satellite data set for the region north of 82°N to examine the factors underlying the model bias in sea ice thickness. There are large differences among the radiation components of the reanalysis products and the model output, with clear anomalies in the radiation of two models. The upward shortwave radiation of FGOALS-f3-L in March and that of CIESM in September are remarkably low, which may be the reason for the bias in the sea ice thickness simulation of these two models.