Water resource management is often based on numerical models, and large-scale models are sometimes used for international strategic agreements. Sometimes the modelled area entails several political entities and river basins. To avoid methodological bias in results, methods and databases should be homogenous across political and geophysical boundaries, but this may involve fewer details and more assumptions. This paper quantifies the uncertainty when the same model code is applied using two different input datasets; a more detailed one for the country of Sweden (S-HYPE) and a more general one for the entire Baltic Sea basin (Balt-HYPE). Results from the two model applications were compared for the Swedish landmass and for two specific Swedish river basins. The results show that both model applications may be useful in providing spatial information of water and nutrients at various scales. For water discharge, most relative errors are <10% for S-HYPE and <25% for Balt-HYPE. Both applications reproduced the most mean concentration for nitrogen within 25% of the observed mean values, but phosphorus showed a larger scatter. Differences in model set-up were reflected in the simulation of both spatial and temporal dynamics. The most sensitive data were precipitation/temperature, agriculture and model parameter values.