Abstract: The Coupled Model Intercomparison Project phase six (CMIP6) organized by the World Climate Research Project (WCRP) is currently on-going. We compared the outputs from 9 earth climate system models that are developed in China with the observations and evaluated the simulation of the spatial distribution of Arctic sea ice concentration and its long-term trend from 1980 to 2014. It shows that all models perform well in simulating the distribution of sea ice in the Arctic basin in March. The errors exist mainly in marginal areas around the sea ice cover, with maximum errors being in the central Sea of Okhotsk and the Barents Sea, reaching 90%. However, the simulated sea ice concentration in September has above 15% errors in both the marginal areas around sea ice cover and the central part of the Arctic basin. In terms of the spatial distribution of the long-term trend of sea ice concentration, the 9 models overestimate the declining of the area of sea ice concentration in March, with large errors (over 50%) existing in the Sea of Okhotsk, the Barents Sea, and northern area to the Greenland. Compared with the situation in March, the models tend to have larger errors in the area and magnitude of the sea ice declining in September. In addition, the skill of the models on simulating the seasonality of sea ice concentration is related to their performance on simulating the long-term trend of sea ice concentration. The models that can well reproduce the seasonality of sea ice concentration usually performs better on simulating the long-term trend. New parameterization schemes in sea ice model show great potential in improving the performance of climate system models.