Assessing uncertainties in hydrologic models can improve accuracy in predicting future streamflow. Here, simulated streamflows using the Variable Infiltration Capacity (VIC) model at coarse (°) and fine (°) spatial resolutions were evaluated against observed streamflows from 217 watersheds. In particular, the adequacy of VIC simulations in groundwater- versus runoff-dominated watersheds using a range of flow metrics relevant for water supply and aquatic habitat was examined. These flow metrics were 1) total annual streamflow; 2) total fall, winter, spring, and summer season streamflows; and 3) 5th, 25th, 50th, 75th, and 95th flow percentiles. The effect of climate on model performance was also evaluated by comparing the observed and simulated streamflow sensitivities to temperature and precipitation. Model performance was evaluated using four quantitative statistics: nonparametric rank correlation ρ, normalized Nash–Sutcliffe efficiency NNSE, root-mean-square error RMSE, and percent bias PBIAS. The VIC model captured the sensitivity of streamflow for temperature better than for precipitation and was in poor agreement with the corresponding temperature and precipitation sensitivities derived from observed streamflow. The model was able to capture the hydrologic behavior of the study watersheds with reasonable accuracy. Both total streamflow and flow percentiles, however, are subject to strong systematic model bias. For example, summer streamflows were underpredicted (PBIAS = −13%) in groundwater-dominated watersheds and overpredicted (PBIAS = 48%) in runoff-dominated watersheds. Similarly, the 5th flow percentile was underpredicted (PBIAS = −51%) in groundwater-dominated watersheds and overpredicted (PBIAS = 19%) in runoff-dominated watersheds. These results provide a foundation for improving model parameterization and calibration in ungauged basins.