Permafrost is one of the most important components of Arctic land. Regional atmosphere-snowpermafrost interactions can be best studied with Regional Climate Models (RCMs) due to their higher horizontal resolution compared to global climate models. The development of Arctic RCMs with sophisticated land models is therefore very important. Comparing RCMs with different land surface model (LSM) components then allows the quantification of the uncertainties associated with the LSM. This study analyzes two simulations of coupled atmosphere-land RCMs over the Arctic, which differ only in their land component, while the atmospheric model component is the same. Specifically, we examine HIRHAM5-CLM4 (HIRHAM5 coupled with the sophisticated land model CLM4) and HIRHAM5 (HIRHAM5 coupled with the simpler land model of ECHAM5). We discuss the two models’ abilities to represent observations on permafrost-like permafrost extent, active layer thickness (ALT), and soil temperature profiles, as well as on the representation of the Arctic atmosphere, based on simulations over 1979–2014. In comparison to HIRHAM5, HIRHAM5-CLM4 significantly reduces the simulated bias in ALT and winter soil temperatures. We find that the simulation of soil temperature and subsequently ALT is sensitive to soil thermal and hydraulic parameter representation in the models. The simulation of permafrost extent is sensitive to the initial soil temperature state in the models. Both HIRHAM5 and HIRHAM5-CLM4 do similarly well in modeling the Arctic 2 m air temperature and atmospheric circulation. Changing the land model impacts the 2 m air temperature significantly over land and the atmospheric circulation predominantly over the Arctic Ocean, associated with changes in baroclinic cyclones.