An observed daily peak wind gusts (DPWG) dataset over Scandinavia, consisting of time series from 127 meteorological stations across Finland, Norway and Sweden, has been created. This dataset provides high‐quality and homogenized near‐surface DPWG series for Scandinavia, spanning the longest available time period (1996–2016). The aim of this study is to evaluate the ability of two regional climate models (RCMs) in simulating DPWG winds. According to the observed DPWG climatology, meteorological stations are classified into three regions for which wind conditions are influenced by similar physical processes: coast, inland and mountain. Smaller‐scale DPWG features of the three regions are only captured when coarser general circulation models or reanalyses are downscaled by a RCM. Dynamic downscaling is thus needed to achieve more realistic simulations of DPWG when compared to their driving models. The performances of the RCMs are found to be more dependent on model dynamics and physics (such as gust parametrization) than on the boundary conditions provided by the driving models. We also found that the RCMs cannot accurately simulate observed DPWG in inland and mountainous areas, suggesting the need for higher horizontal resolution and/or better representation of relevant boundary‐layer processes.