About 600 ozone sonde profiles and in situ ozone observations on 2300 aircraft flights were compared with several three-dimensional global chemistry-transport models, in the midlatitude lowermost stratosphere during 1996. The models use the same top boundary conditions and parameterized ozone chemistry. The comparisons show that the models using general circulation model (GCM) winds do not capture the seasonal ozone accumulation in the lowermost stratosphere. The models using winds from the European Center for Medium-Range Weather Forecasts (ECWMF) overestimate ozone above this maximum, especially during spring. This overestimation increases with increasing latitude. Close to the tropopause, the best agreement between models and observations is found during winter and the worst occurs during summer, where most models underestimate the mean ozone concentration. This underestimation is partly caused by inaccurate description of the relatively small-scale transport processes, mainly associated with convective activity, and partly by incomplete ozone chemistry in the parameterized scheme. The models that use assimilated winds show significant differences in calculated ozone, despite their common source (ECWMF) to calculate the transport. We illustrate that the model performance significantly depends on the ECWMF processing method and that interpolation of wind data should be avoided. In addition, the results seem relatively insensitive to vertical resolution. To improve the model performance further in this region, the horizontal resolution should be higher than 2.5¿. ¿ 2001 American Geophysical Union