Ozone loss during the winter 2002/2003 has been evaluated from comparisons between total ozone reported by the SAOZ network and simulated in passive mode by both REPROBUS and SLIMCAT. Despite the fact that the two models have a different approach to calculate the descent inside vortex, both evaluations provide similar results 18±4% using REPROBUS and 20±4% using SLIMCAT and show that the loss started around mid-December, at least ten to twenty days earlier than during any of the previous eleven winters, except 1993/1994. This unusual behaviour is consistent with the low temperatures reported in the stratosphere as well to the signature of early chlorine activation indicated by ground-based, balloon and satellite observations. A significant ozone l oss is also simulated by the current versions of two models, but of lesser amplitude compared to SAOZ, 13±2% for REPROBUS and 16±2% for SLIMCAT, the underestimation being already observed by mid-January. The early ozone depletion captured by both model show that chemical depletion did indeed take place in December, predominantly at the illuminated edge of the distorted vortex, but the reason for the underestimation compared to the observations and the differences among the models have still to be investigated.