The main aims of this paper are the evaluation of five large-scale hydrological models across Europe and the assessment of the suitability of the models for making projections under climate change. For the evaluation, 22 years of discharge measurements from 46 large catchments were exploited. In the reference simulations forcing was taken from the E-OBS dataset for precipitation and temperature, and from the WFDEI dataset for other variables. On average across all catchments, biases were small for four of the models, ranging between −29 and +23 mm yr−1 (−9 and +8 %), while one model produced a large negative bias (−117 mm yr−1; −38 %). Despite large differences in e.g. the evapotranspiration schemes, the skill to simulate interannual variability did not differ much between the models, which can be ascribed to the dominant effect of interannual variation in precipitation on interannual variation in discharge. Assuming that the skill of a model to simulate interannual variability provides a measure for the model's ability to make projections under climate change, the skill of future discharge projections will not differ much between models. The quality of the simulation of the mean annual cycles, and low and high discharge was found to be related to the degree of calibration of the models, with the more calibrated models outperforming the crudely and non-calibrated models. The sensitivity to forcing was investigated by carrying out alternative simulations with all forcing variables from WFDEI, which increased biases by between +66 and +85 mm yr−1 (21–28 %), significantly changed the inter-model ranking of the skill to simulate the mean and increased the magnitude of interannual variability by 28 %, on average.