Year-long measurements of NO x and ozone performed during the NOXAR project are compared to results from the ECHAM4.L39(DLR)/CHEM (E39/C) and GISS coupled chemistry-climate models. The measurements were taken on flights between Europe and the eastern United States and between Europe and the Far East in the latitude range 40-65°N. Our comparison concentrates on the upper troposphere and reveals strong longitudinal variations in seasonal mean NO x of more than 200 pptv, which both models are able to reproduce qualitatively. Vertical profiles show maximum NO x values 2-3 km below the tropopause ("E-shape") with a strong seasonal cycle. E39/C simulates a maximum located at the tropopause and with a reasonable seasonal cycle. The GISS model reproduces the seasonal cycle but not the profile's shape due to its coarser vertical resolution. A comparison of NO x frequency distributions reveals that both models are capable of reproducing the observed variability, except that E39/C shows no very high NO x mixing ratios. Both models show that lightning and surface NO x emissions contribute the most to the seasonal cycle of NO x at tropopause altitudes. The impact of lightning in the upper troposphere does not vary strongly with altitude, whereas the impact of surface emissions decreases with altitude. Among all sources, lightning contributes the most to the variability of NO x in the upper troposphere in northern mid-latitudes during summer.