Abstract: Based on various datasets, such as CMA-GFS (Global Forecast System of China Meteorological Administration), global sounding observations, ground-encrypted automatic station observations, and ERA5 reanalysis results, 37 cold wave events from 2000 to 2024 were screened, and the distribution characteristics of cold wave cooling amplitude and wind speed in the Beijing–Tianjin–Hebei (BTH) region were statistically analyzed. Next, a multidimensional test was conducted to investigate the spatial and temporal deviations of CMA-GFS forecasts for high-altitude variables, the highest and lowest ground temperatures at 2-m height, and the wind field at 10-m height. These parameters were tested across different forecast lead times (0–72 h). The associated multifactor forecasting performance was carefully evaluated, and possible causes of the forecasting bias of the ground-temperature model were discussed. Results reveal the following: (1) regions with an average maximum cooling amplitude more than 6°C during cold wave events are concentrated in the northern and southern parts of Beijing, Tianjin, and southeastern Hebei. The cooling amplitude in Beijing’s northern mountainous areas is significantly higher than that in the plain areas. Additionally, areas with a maximum cooling amplitude more than 16°C are mainly located in mountainous regions, whereas the maximum cooling amplitude at most other stations is concentrated at 8°C–12°C. The distributions of 2 min average wind speed and maximum gust values are similar, with overall lower and higher wind speeds in noncoastal plain areas and mountainous and coastal areas, respectively. (2) CMA-GFS can effectively predict the atmospheric-circulation situation in the upper troposphere during cold wave events. However, its temperature predictions exhibit certain deviations. For most events, deviations do not decrease with shorter forecast lead times, and some events show the largest prediction deviation at the shortest lead time. Notably, forecasts of the highest ground temperature are better than those of the lowest. Further, the highest ground temperature is underestimated in 63.6% of cold wave events, with areas with low forecast values concentrated in Beijing, Tianjin, and central Hebei. The mean error, root mean square error, station accuracy, and deviation dispersion reported at 0-h forecast lead time are better than those at other lead times. For most cold wave events, the lowest ground temperature is overestimated, particularly in high-altitude areas in northern and western of Hebei and Northern Beijing. The best performance is realized at 36-h forecast lead time, and the maximum error is achieved at 0-h forecast lead time. The 10-m wind speed prediction error is within ±2 m s⁻¹ for most stations in the BTH region, and a few stations are more than 2 m s⁻¹. (3) The deviations in temperature at 10-m height forecast by the CMA-GFS may be related to the inaccuracies in forecasting temperature and humidity in the middle-lower atmospheric layers and the overestimation of the wind speed at 10-m height.