Abstract. Four regional chemistry transport models were applied to simulate the concentration and composition of particulate matter (PM) in Europe for 2005 with horizontal resolution  ∼  20 km. The modelled concentrations were compared with the measurements of PM chemical composition by the European Monitoring and Evaluation Programme (EMEP) monitoring network. All models systematically underestimated PM₁₀ and PM_2.5 by 10–60 %, depending on the model and the season of the year, when the calculated dry PM mass was compared with the measurements. The average water content at laboratory conditions was estimated between 5 and 20 % for PM_2.5 and between 10 and 25 % for PM₁₀. For majority of the PM chemical components, the relative underestimation was smaller than it was for total PM, exceptions being the carbonaceous particles and mineral dust. Some species, such as sea salt and NO₃⁻, were overpredicted by the models. There were notable differences between the models' predictions of the seasonal variations of PM, mainly attributable to different treatments or omission of some source categories and aerosol processes. Benzo(a)pyrene concentrations were overestimated by all the models over the whole year. The study stresses the importance of improving the models' skill in simulating mineral dust and carbonaceous compounds, necessity for high-quality emissions from wildland fires, as well as the need for an explicit consideration of aerosol water content in model–measurement comparison.