By using the hindcast and forecast data from the National Centers for Environmental Prediction Climate Forecast System version 2 (NCEP CFSv2) for the 1982–2018 period, we investigate the forecasting skills of the mid‐summer (July and August) surface air temperature (SAT) at interannual timescales in this study. Although CFSv2 predictions show a warm bias for the climatological mean SAT over the Yangtze River valley (25°–32°N, 105°–122°E), they show a consistent and great performance in predicting the interannual variability of the mid‐summer SAT over this region until 4 months in advance, where the linear correlation coefficient between the predicted and observed time series reaches +0.65, +0.51 and + 0.68 for 4, 2 and 0 months in advance, respectively. The CFSv2 predictions well simulate the linkage between the SAT anomalies over the Yangtze River valley and the anomalous atmospheric circulation aloft, including the circumglobal teleconnection and zonal extension of the Western Pacific Subtropical High. However, CFSv2 has trouble in simulating the associated vertical velocity, cloud cover and solar radiation anomalies, except for 0 months in advance. The persistent forecasting skills result from the accurate response of the circumglobal teleconnection and Western Pacific Subtropical High to the El Niño/Southern Oscillation and sea surface temperature (SST) anomalies over the mid‐latitude North Atlantic. Correspondingly, the forecasting skill, signal and signal‐to‐noise ratio are effectively improved in the years with strong mid‐summer SST anomalies over the tropical central‐eastern Pacific or mid‐latitude North Atlantic. These results are useful for understanding the predictability of the mid‐summer SAT over the Yangtze River valley.