Long‐term hydrological forecasting, water resources management and other climate change impacts or adaptation analysis studies on large continental river basins, for example, the Athabasca River Basin (
ARB) in Canada, desire a reliable climatic projection. This usually relies on general circulation models ( GCMs) in the fifth phase of the Coupled Model Intercomparison Project ( CMIP5). However, there is a lack of a systematic evaluation of CMIP5 GCMperformances over the ARBthat vary with multiple factors, for example, statistical metrics, temporal scales and spatial locations, challenging the reliability of water‐related or other studies over the ARB. For this gap to be filled, six CMIP5 GCMs, namely, IPSL‐CM5A‐LR, IPSL‐CM5A‐MR, MIROC‐ESM‐CHEM, MIROC5, GFDL‐ESM2Gand GFDL‐ESM2M, and their ensemble mean are selected according to data availabilities of representative climate variables: Tmin, Tmaxand Prec( TTP). Accuracies of the selected CMIP5 GCMs in reproducing TTPover the ARBare evaluated comprehensively. The ensemble mean cannot outperform any GCMin all cases in the ARB, although its overall accuracy seems to be higher in consideration of all cases. These accuracies vary with TTP, locations, metrics and scales. For instance, ESM2Gshows the highest accuracies in reproducing monthly/seasonal variability and magnitudes of grid‐averaged TTPand inter‐annual variability of grid‐averaged annual means of Tmax; CM5A‐LRin multi‐year‐averaged spatial variability of TTPand magnitudes of spatially distributed multi‐year‐averaged Tmax; while the ensemble mean only in some aspects, for example, intraseasonal variability and magnitudes of TTPand inter‐annual variability and magnitudes of grid‐averaged annual means of TTP. GCMs should be systematically integrated according to accuracy variations. Multiple statistical metrics are recommended in GCMevaluations. These findings facilitate water resources systems analyses and other related studies in the ARB. Copyright © 2017 John Wiley & Sons, Ltd.