Nitrogen and phosphorus retention estimates in streams and standing water bodies were compared for four European catchments by a series of catchment-scale modelling tools of different complexity, ranging from a simple, equilibrium input-output type to dynamic, physical-based models: source apportionment, MONERIS, EveNFlow, TRK, SWAT, and NL-CAT. The four catchments represent diverse climate, hydrology, and nutrient loads from diffuse and point sources in Norway, the UK, Italy, and the Czech Republic. The models' retention values varied largely, with tendencies towards higher scatters for phosphorus than for nitrogen, and for catchments with lakes (Vansjo-Hobol, Z. elivka) compared to mostly or entirely lakeless catchments (Ouse or Enza, respectively). A comparison of retention values with the size of nutrient sources showed that the modelled nutrient export from diffuse sources was directly proportional to retention estimates, hence implying that the uncertainty in quantification of diffuse catchment sources of nutrients was also related to the uncertainty in nutrient retention determination. This study demonstrates that realistic modelling of nutrient export from large catchments is very difficult without a certain level of measured data. In particular, even complex process oriented models require information on the retention capabilities of water bodies within the receiving surface water system and on the nutrient export from micro-catchments representing the major types of diffuse sources to surface waters.