Abstract | In recognition of the large capacity and demand uncertainties involved in the seismic response of buildings, the engineering community has increasingly become, over the years, convinced that the problem may be treated more efficiently using probabilistic tools coupled with appropriate performance-related criteria. So far, the majority of the work carried out on the reliability of steel Moment Resisting Frames (MRFs) has focused on buildings typical of US design and construction practice, which are, in principle, different from their European counterparts. Thus, this study presents the results of a full probabilistic assessment performed on an EC8-designed, 5-storey steel MRF. In order to obtain fragility curves, the random structural responses were evaluated through a large number of non-linear time history analyses using an ensemble of European ground records. The limit state functions necessary to define fragility were expressed in terms of the previously mentioned structural responses as well as the probabilistically-defined FEMA 356 performance levels. Using these fragilities, annual failure probabilities were obtained for the city of Reggio Calabria in Italy. The aforementioned fragilities were then compared against those obtained from an equivalent SDOF coupled with a closed-form lognormal approximation. The study has revealed that (a) the influence of the ground record characteristics on fragility is significant, (b) the fragility variance increases with the level of induced non-linearity and (c) the fragility methodology based on the equivalent SDOF yields reasonable results.