Air temperature at 2-m (T2) in the Arctic represents its local climate. Its quantification is one of the major criteria to evaluate the performance of numerical models in reflecting the complex physical and dynamical processes associated with the surface energy balance. This study uses HIRHAM5 regional climate model to simulate the Arctic climate during 1979–2014. Evaluations with Arctic station observations reveal that HIRHAM5 can generally reproduce the temporal and spatial variation of the T2, although a systematic cold bias of ca. −2 °C exists in all seasons. The overestimated surface albedo in spring and autumn, and the underestimated downward solar radiation associated with the cloud cover in summer are the main causes of the cold biases in each respective season. The model also simulates the Arctic warming well (with linear trends of 0.40 °C decade−1 for the annual mean T2), although the magnitude is less than that from ERA-Interim (0.55 °C decade−1) and station observations (0.60 °C decade−1). In addition, strong decadal variability is clear in the T2 trends calculated using an 11-year moving windows, especially in winter and spring, which is mainly associated with the variability of the Arctic/North Atlantic Oscillations.