Atmospheric tides play a critical role in the dynamics and coupling of mesosphere and lower-thermosphere (MLT). Global Circulation Models (GCMs) that aim to span the lower, middle and upper atmosphere must therefore be capable of reproducing the tides and observations of tides are thus crucial to constrain the models. Here we present the first climatology of the 12- and 24-h tides measured at heights of 80–100 km by a meteor radar over the Antarctic station of Rothera (68°S, 68°W). We use observations of tides from 2009 in the first test of two GCMs at these latitudes: the Whole Atmosphere Community Climate Model (WACCM) and the extended Canadian Middle Atmosphere Model (eCMAM, 24-h tide only). Our radar observations reveal large-amplitude 12- and 24-h tides which display a distinct seasonal variability. The 12-h tide maximises around the equinoxes, reaching daily-mean amplitudes of about 40 ms−1. The 24-h tide is generally of smaller amplitude and maximises in summer. The observed 12-h tide increases greatly in amplitude with increasing height over the range 80–100 km, whereas the 24-h tide generally does not do so. Comparison with the models shows that, for the 12-h tide, WACCM reproduces the observed amplitudes at heights near 80 km quite well, but does not reproduce either the strong observed increase with height or the equinoctial maxima. For the 24-h tide, WACCM reproduces the observed small variation in amplitude with height, but suggests amplitudes somewhat larger than those observed whilst eCMAM generally reproduces the observed tidal amplitudes and the small variation of amplitude with height. The radar observations reveal great day-to-day variability in the amplitude of both tides, much of which is quasi periodic and occurs at periods similar to those of planetary waves, suggesting that it originates in non-linear tidal/planetary-wave coupling. The observed and model background winds display some notable differences, particularly in winter when eastward winds are observed at all heights but not reproduced in the models. We propose that these differences may arise from the lack of in-situ gravity-wave sources in the models and that this may also account for some of the differences apparent between the observed and modelled tides.