An ensemble of three Regional Climate Models (RCMs) is evaluated over the current climate conditions with the aim of assessing RCMs’ skills and limitations in reproducing the near-surface temperature and precipitation over the Subtropical Chile complex terrain region (25 ºS–45 ºS). The simulations driven by ERA-Interim and by GCMs historical scenarios are compared against observational gridded products and ERA5 reanalysis for high and low elevations separately. The RCMs simulate reasonably well the main spatio-temporal characteristics of temperature and precipitation, such as latitudinal climate gradients, orographic uplifts, phase of the seasonal cycle, and realistic inter-annual variability. Although the simulations driven by ERA-Interim show better skills than those driven by GCMs, especially in the case of precipitation, none of the RCMs/simulations performs best or worst in all sub-regions. RCMs and ERA5 have a common prominent cold bias at high elevations north of 35 °S, which is particularly strong above 2000 m.a.s.l. This bias is associated with a strong overestimation of precipitation and an overestimation of surface albedo, likely related to an overestimated snow cover. Because of scarce in-situ observations, this region is also associated with inherent observational uncertainty. Our results emphasize the necessity of improving the density and quality of observational networks over the complex terrain of Subtropical Chile for an accurate assessment and tuning/calibration of RCMs. The assessment of a state-of-the-art RCMs ensemble provided by this study can help in interpreting, bias-correcting, and using the regional climate projections performed with these RCMs for downstream applications and impact studies.