A coordinated set of large ensemble atmosphere-only simulations is used to investigate the impacts of observed Arctic sea ice-driven variability (SIDV) on the atmospheric circulation during 1979-2014. The experimental protocol permits separating Arctic SIDV from internal variability and variability driven by other forcings including sea surface temperature and greenhouse gases. The geographic pattern of SIDV is consistent across seven participating models, but its magnitude strongly depends on ensemble size. Based on 130 members, winter SIDV is~0.18 hPa 2 for Arctic-averaged sea level pressure (~1.5% of the total variance), and~0.35 K 2 for surface air temperature (~21%) at interannual and longer timescales. The results suggest that more than 100 (40) members are needed to separate Arctic SIDV from other components for dynamical (thermodynamical) variables, and insufficient ensemble size always leads to overestimation of SIDV. Nevertheless, SIDV is 0.75-1.5 times as large as the variability driven by other forcings over northern Eurasia and Arctic. Plain Language Summary Changing Arctic sea ice conditions since the late 1970s have exerted profound impacts on environment and ecosystem at the high latitudes and have been suggested to affect midlatitude weather and climate, although this topic has been controversial. In order to improve our understanding on how Arctic sea ice changes influence local and remote weather and climate, a coordinated set of experiments has been performed using various state-of-the-art atmosphere-only models to study the linkages between the Arctic climate change and lower latitudes. This study uses seven models following a common experimental protocol to investigate the atmospheric circulation changes forced by Arctic sea ice variability. The protocol allows the Arctic sea ice-driven variability (SIDV) to be singled out. In boreal winter, the Arctic SIDV is~0.18 hPa 2 and accounts for only~1.5% of the total variance for sea level pressure, while it is~0.35 K 2 and accounts for~21% for surface air temperature. The results also suggest that using insufficient ensembles always leads to an overestimation of SIDV, and more than 100 and 40 ensemble members are needed for sea level pressure and surface air temperature within the Arctic, respectively, to separate the SIDV from the variability due to other factors, primarily the atmospheric internal variability.