The reliability of climate simulations and projections, particularly in the regions with complex terrains, is greatly limited by the model resolution. In this study, we evaluate the variable-resolution Community Earth System Model (VR-CESM) with a high-resolution (0.125 deg) refinement over the Rocky Mountain region. The VR-CESM results are compared with observations, as well as CESM simulation with a quasi-uniform 1 deg resolution (UNIF) and Canadian Regional Climate Model version 5 (CRCM5) simulation at a 0.11 deg resolution. We find that VR-CESM is effective at capturing the observed spatial patterns of temperature, precipitation, and snowpack in the Rocky Mountains with the performance comparable to CRCM5, while UNIF is unable to do so. VR-CESM and CRCM5 capture better the seasonal variations of precipitation than UNIF, but VR-CESM overestimates winter precipitation whereas CRCM5 and UNIF underestimate it. However, all simulations distribute more winter precipitation along the windward (west) flanks of mountain ridges with the greatest overestimation in VR-CESM. VR-CESM simulates much greater snow water equivalent peaks than CRCM5 and UNIF, although results are still 10-40% less than observations. Moreover, the frequency of heavy precipitation events (daily precipitation≥ 25 mm) in VR-CESM and CRCM5 is comparable to observations, whereas the same events in UNIF are an order of magnitude less frequent. In addition, VR-CESM captures the observed occurrence frequency and seasonal variation of rain-on-snow (ROS) days and performs better than both VR and CRCM5. These results demonstrate the VR-CESM's capability in regional climate modeling over the mountainous regions and its promising applications for climate change studies.