Operational model analyses of the European Centre for Medium-Range Weather Forecast (ECMWF), the German Weather Service (DWD) and the Norwegian Meteorological Institute (HIRLAM) are compared to data measured by 16 ice buoys and by a research aircraft in eight flight missions during the Fram Strait cyclone experiment FRAMZY in March/April 2007. The sea level pressure (SLP) compares well with the buoys for all models, SLP correlations are >0.99, SLP tendency correlations are >0.95 and spatial SLP correlations are >0.94 on the average. The aircraft measurements are used for comparison of the horizontal and vertical boundary layer structure in ECMWF and HIRLAM under cyclonic conditions. Horizontal SLP gradients are slightly underestimated. The surface air temperature (SAT) is too low on the warm side and too high on the cold side of the synoptic systems. The temperature inversion on the cold side is not realistically captured with respect to its base and strength in the ECMWF model and with respect to its thickness in HIRLAM. This results in a wrong vertical structure of humidity and wind. The SAT errors hint at model deficits in the representation of the surface energy balance and the inversion errors hint at deficits in the parametrization of the vertical mixing.