This study assesses the performance of the sixth-generation Canadian Regional Climate Model (CRCM6) in simulating the amount and phase of cold-season precipitation, as well as 2-m air temperature. It also examines the added value of finer grid spacing, which enables the explicit representation of deep convection. Simulations were conducted at grid spacings of 0.11° (» 12 km) and 0.0225° (» 2.5 km), and the results were compared with surface observations from 35 hydrometeorological stations across Quebec over two cold seasons (October to April in 2020–21 and 2021–22). The analysis was further supported by atmospheric sounding data from two stations, and several gridded reference products. Both simulations exhibited similar large-scale 2-m air temperature spatial patterns, with the coarser-resolution simulation generating consistently colder values, as well as a larger total precipitation amount and bias compared to station observations. The finer-resolution simulation reduced the total precipitation bias by a factor of three. Both simulations overestimated liquid precipitation and underestimated solid precipitation, with the finer resolution better capturing liquid precipitation and the coarser resolution better capturing solid precipitation. Mixed precipitation remained a challenge, being simulated at colder temperatures than observed, particularly near 0°C. The 50% rain-snow temperature threshold (T50), which indicates when liquid and solid phases occur equally, was 0.5°C for the finer resolution and 0.8°C for the coarser resolution, both below the observed 2.1°C. This study highlights the need to refine the model’s representation of mixed-phase precipitation and rain-snow transitions.