Sub-seasonal to seasonal (S2S) climate and runoff prediction is of great importance for active disaster reduction. The prediction models of runoff anomaly and extreme drought for the future 40 days are developed based on CMA-CPS v3 climate model by National Climate Center, China Meteorological Administration, and a hydrological model HBV. The performance of the models are evaluated with the indices of MSSS, ACC, AUC and BSS for hindcast, and verified for the runoff anomaly prediction over the Yellow River Basin in 2021. The results suggest that runoff mean prediction for the future 40 days is skillful, and the skill is higher in dry season than wet season. Moreover, the extreme drought prediction is skillful in later autumn (November) and the winter months (December, January and February). The direction of monthly runoff anomaly from May to October in 2021 are predicted correctly except in June and September, but the anomaly degrees are different from observation. Finally, the analysis of skill variation with lead time and seasons, and the skill difference between runoff and precipitation reveals that the skill of precipitation prediction by S2S climate model influences runoff skill, especially during wet season. However, there are also other factors than precipitation affecting the skill of runoff prediction.