Abstract: There are 52 models providing simulations of historical sea ice concentration and thickness in the CMIP6, and 35 of them provide future sea ice simulations under moderate forcing scenario (SSP245) and high forcing scenario (SSP585). In this paper, the 52 models are comprehensively evaluated on the simulation of sea ice in March, September and seasonal variation. The 35 models were further used to conduct sea ice extent projection tests from 1982 to 2012, and the top 19 models were selected to predict the future Arctic summer sea ice free time when sea ice extent is less than one million square kilometers. Under the moderate forcing scenario (SSP245), the result from the ensemble regression method indicates that the Arctic will be ice free in summer for the first time in 2065 with sea ice extent less than one million square kilometers, while the result from the multi-model ensemble mean method shows that the ice free in summer will occur in 2064. The time difference between the two projections is only one year. Under the high forcing scenario (SSP585), the ensemble regression method indicates that Arctic will be ice free in summer for the first time in 2051, while the multi-model ensemble mean method shows that the ice free in summer will occur in 2050. The time difference is also only one year. When using the same model simulations, the projections of the two methods have little difference. The uncertainty of the above projections still exists as in pervious studies.