AbstractReliable predictions of oceanic waves during storms have always been foremost for offshore design and operation, coastal hazards, and navigation safety. Indeed, many accidents that occurred during storms were ascribed to the impact with unforeseen large waves. In this context, the purpose of this study is to improve the present state extreme wave estimate from spectral wave models. We describe an implementation for the WAM model, and we investigate the use of WAM and WAVEWATCH III fed with common routines designed to evaluate the short-term/range maximum wave statistics. An extensive assessment of models' results in the Adriatic and North Sea is performed using time and space-time wave measurements, and through an intercomparison between WAM and WAVEWATCH III applied with three different input/dissipation source term parametrizations (ST3/4/6). Further, models’ capabilities are investigated, and extreme waves characterized, in the Mediterranean Sea, aiming also at disentangling the wave spectrum bulk parameters that may point to favorable conditions for the generation of high waves. Based on the comparisons between model results and measurements, we conclude that for the model characterization of extremes, the accuracy of the significant wave height is pivotal; differences between models of other spectral parameters seem to have a minor effect.