A method for verifying precipitation predication of transition weather is developed. It focuses on the evaluation of the performance of numerical prediction models from the prediction capability for transition weather. The precipitation predictions from several global medium range numerical prediction models (including China, Japan and German) are selected and verified by using this method during the period from September 2006 to August 2008. The verification results show that the method is an effective complement to the currently used precipitation prediction verification methods. The forecast skills for transition weather tend to decrease with the growth of leadtime. The short range forecast skill analysis indicates that T213 and JMA global models perform better in spring and the German model is the best in summer. But for 48 hour prediction, results are different, depending on the forecast areas. For example, T213 and the JMA model perform better in the middle lower reaches of the Yangtze River and in some parts of North and Northeast China. The German model performs best in Sichuan Basin and South China.