Mechanistic model simulations of fall/early winter in the northern (November and December) and southern (May and June) stratosphere are compared with observational analyses to examine the skill of the model in simulating the state of the stratosphere, including both means and variability in key fields, during six winters. While detailed success varies from year to year, the model produces a realistic climatology of and variability in the evolution of winds, geopotential heights, temperatures and wave propagation in the early-winter stratosphere. The variability and mean fields agree well with those in longer data records. The northern hemisphere (NH) simulations show a small cold bias when averaged over the 6 years, while the southern hemisphere (SH) simulations show a larger warm bias. Greater detailed success in simulations of alternate NH winters suggests the possibility of greater model skill during the westerly quasi-biennial oscillation phase, although the short record and complexity of interactions between tropics and high latitudes preclude definitive identification of such a relationship. A prominent failing of the model when using Rayleigh friction to parametrize gravity-wave drag is an inability to correctly reproduce the latitudinal structure of the stratospheric jet above about 7 hPa; this failing can be alleviated by using a non-orographic gravity-wave drag parametrization, at the expense of frequently degrading agreement of planetary-wave phases and amplitudes. The success of the model in reproducing realistic climatology and variability makes these simulations useful for more detailed studies of transport and vortex evolution in early winter. Interhemispheric comparisons show that the early-winter circulations are qualitatively similar in the NH and SH: they are dominated by a strengthening vortex, with most wave activity being in wave 1; they both have minor warmings in a preferred location and clustered around a preferred time—early December (early June) in the NH (SH); variability in the flow shows a crescent pattern of maximum variations near 60–80° latitude centred at the location of the minor warmings (over the dateline in the NH and in the South Pacific in the SH). Interhemispheric differences are primarily in the magnitude of the variability, with a more symmetric circulation and weaker minor warmings in the SH; this is in contrast to later in the winter, when large qualitative interhemispheric differences have been seen. Copyright © 2002 Royal Meteorological Society