Earth system models (ESMs) can help to improve the understanding of climate-induced cryospheric–hydrological impacts in complex mountain regions, such as High Mountain Asia (HMA). Coarse ESM grids, however, have difficulties in representing cryospheric–hydrological processes that vary over short distances in complex mountainous environments. Variable-resolution (VR) ESMs can help to overcome these limitations through targeted grid refinement. This study investigates the ability of the VR Community Earth System Model (VR-CESM) to simulate cryospheric–hydrological variables such as the glacier surface mass balance (SMB) over HMA. To this end, a new VR grid is generated, with a regional grid refinement up to 7 km over HMA. Two coupled atmosphere–land simulations are run for the period 1979–1998. The second simulation is performed with an updated glacier cover dataset and includes snow and glacier model modifications. Comparisons are made to gridded outputs derived from a globally uniform 1∘ CESM grid, observation-, reanalysis-, and satellite-based datasets, and a glacier model forced by a regional climate model (RCM). Climatological biases are generally reduced compared to the coarse-resolution CESM grid, but the glacier SMB is too negative relative to observation-based glaciological and geodetic mass balances, as well as the RCM-forced glacier model output. In the second simulation, the SMB is improved but is still underestimated due to cloud cover and temperature biases, missing model physics, and incomplete land–atmosphere coupling. The outcomes suggest that VR-CESM could be a useful tool to simulate cryospheric–hydrological variables and to study climate change in mountainous environments, but further developments are needed to better simulate the SMB of mountain glaciers.