<p>Gridded precipitation datasets are increasingly used as alternatives to sparse gauged observations for hydrological modelling in topographically complex regions such as the Himalayas. However, their suitability is often assessed only through statistical comparisons with available gauged data, which may not adequately reflect their functional reliability for hydrological modelling. To address this, we evaluated seven gridded precipitation datasets, including APHRODITE, CHIRPS, GPM, PERSIANN, ERA5, ERA5-LAND, and IMDAA, over the Himalayan Upper Beas catchment both statistically and hydrologically. Statistical evaluation compared the precipitation datasets against the India Meteorological Department precipitation from 2001 to 20. Capabilities of detecting the precipitation events and estimating the precipitation magnitudes were assessed using categorical metrics (POD, FAR, and CSI) and statistical metrics (CC, R², RMSE, and RB) respectively. Statistical evaluation showed that APHRODITE, ERA5-LAND, ERA5, GPM, and PERSIANN exhibited satisfactory performance. Further, hydrological evaluation compared the streamflow simulations, obtained using the precipitation datasets as input to a hydrological model like SWAT, against the observed streamflow from 2014 to 20. Results revealed substantial divergence from the statistical evaluation outcomes. Although multiple datasets performed well statistically, only APHRODITE, PERSIANN, and GPM satisfactorily reproduced observed streamflow (NSE = 0.88, 0.65, 0.61; RSR = 0.34, 0.59, 0.62; and PBIAS = -2.18%, -10.86%, -18.98%). It provides empirical evidence that reliance on statistical evaluation alone may lead to misleading conclusions regarding the suitability of gridded precipitation datasets for hydrological modelling. The study, therefore, highlights the necessity of a combined statistical and hydrological evaluation framework to identify functionally reliable precipitation datasets for hydrological applications in complex mountainous regions.</p>