A regional climate model of the whole Arctic using the dynamical package of the High-Resolution Limited Area Model (HIRLAM) and the physical parameterizations of the Hamburg General Circulation Model (ECHAM3) has been applied to simulate the climate of the Arctic north of 65 ° N at a 50-km horizontal resolution. The model has been forced by the European Centre for Medium-Range Weather Forecasts (ECMWF) analyses at the lateral boundaries and with climatological or actual observed sea surface temperatures and sea ice cover at the lower boundary. The results of simulating the Arctic climate of the troposphere and lower stratosphere for January 1991 and July 1990 have been described. In both months the model rather closely reproduces the observed monthly mean circulation. While the general spatial patterns of surface air temperature, mean sea level pressure, and geopotential are consistent with the ECMWF analyses, the model shows biases when the results are examined in detail. The largest biases appear during winter in the planetary boundary layer and at the surface. The underestimated vertical heat and humidity transport in the model indicates the necessity of improvements in the parameterizations of vertical transfer due to boundary layer processes. The tropospheric differences between model simulations and analyses decrease with increasing height. The temperature bias in the planetary boundary layer can be reduced by increasing the model sea ice thickness. The use of actual observed sea surface temperatures and sea ice cover leads only to small improvements of the model bias in comparison with climatological sea surface temperatures and sea ice cover. The validation of model computed geopotential, radiative fluxes, surface sensible and latent heat fluxes and clouds against selected station data shows deviations between model simulations and observations due to shortcomings of the model. This first validation indicates that improvements in the physical parameterization packages of radiation and in the description of sea ice thickness and sea ice fraction are necessary to reduce the model bias.