A model intercomparison between two atmospheric models, the non-hydrostatic Lokal Modell (LM) and the hydrostatic HIgh Resolution Limited Area Model (HIRLAM) is carried out for a one-week period, including a case of cyclogeneis leading to heavy precipitation over Northern Italy. The two models, very different in terms of data-assimilation and numerics, provide different results in terms of forecasts of surface fields. Opposite diurnal biases for the two models are found in terms of screen level temperatures. HIRLAM wind speed forecasts are too strong, while LM precipitation forecasts have larger extremes. The intercomparison exercise identifies some systematic differences in the weather products generated by the two systems and sheds some light on the biases of the two numerical weather prediction systems.