An assessment is presented of the extent and variability of Antarctic sea ice in the non-flux-corrected version of the Hadley Centre's coupled atmosphere-ocean general circulation model (HadCM3). The results arc based on a 100 year segment of a long control run of the model with the sea ice being compared to ice extents and concentrations derived from passive microwave satellite data. Over the year as a whole, the model ice extent (the area with > 15 % ice concentration) is 91 % of that determined from satellite imagery, but, not surprisingly, the regional-scale distribution differs from the observed. Throughout the year there is too much ice near 90degrees E, which is believed to be present as a result of incorrect ocean currents near Kerguelen. In contrast to the satellite data, there is too little ice to the west of the Antarctic Peninsula as a result of anomalously northerly atmospheric flow, compared to observations. During the winter the sea-ice concentrations in the model are too high, possibly as a result of the simple representation of the sea ice, which does not simulate complex dynamical interactions within the pack, The annual cycle of sea-ice advance/retreat in the model has a phase error, with the winter sea-ice maximum extent being too late by about 1 month.