Abstract: In this study, the performances of the median-resolution (about 110 km) climate system model (BCC-CSM2-MR) and atmospheric model (BCC-AGCM3-MR) developed at the National Climate Center are evaluated using three reanalysis datasets (ERA5, JRA55, and NCEP/NCAR). The results show that a high blocking tendency can be distinguished over the “North Atlantic–Western Europe” and “Central North Pacific”. The winter and spring blocking frequency is almost twice of that in summer and autumn. The blocking frequency in the ERA5 dataset is higher than that in the JRA55 and NCEP/NCAR datasets, especially over the North Pacific. Model evaluations show that the atmospheric model reproduces the main features of the Northern Hemisphere blocking frequency, spatial structures, and seasonal variations very well. The main biases of the model are overestimation over Europe–Asia in winter and spring, especially over the Ural Mountains, and underestimations over the North Atlantic. These biases are attributed to the climatological biases in geopotential heights at 500 hPa. The overall performances of the BCC-CSM2-MR model are similar to those of the BCC-AGCM3-MR model. However, the winter and spring blocking over Europe–Asia and especially over the Ural Mountains, is improved. Spring blocking and the double-peak blocking structure in summer over the North Pacific are also better reproduced in the coupled model. Therefore, air–sea coupling may help to improve the reproduction of blocking frequency over Europe–Asia and the North Pacific. The internal variability of the climate system has a great impact on the variation of the blocking frequency, which also affects the capability of the model to predict blocking highs.