Short- and medium-term precipitation forecast products are important to improve the prediction period and accuracy of flood forecasts. With global climate change, the prediction of precipitation patterns becomes increasingly complex and important. To date, there have been no available reports on the applicability of TIGGE precipitation products (a significant aspect of the most authoritative data set for short-and medium-term ensemble forecasts) in the arid and semi-arid regions of China. On the basis of precipitation data measured from 2015-2017 in arid and semi-arid regions of China, the mean absolute deviation, root mean square error, TS score, and other indicators were used to analyze the precipitation forecast, precipitation classification forecast, precipitation detection ability, and spatial prediction accuracy. The prediction effects of the ECMWF, JMA, KMA, and UKMO models in the TIGGE data center were evaluated comprehensively in the study area. The results show that the four models effectively forecast light rain. The JMA model has the best forecast efficacy for light rain at different precipitation levels; for other precipitation levels, no significant difference was seen between the four models. The KMA model performs the worst for the daily precipitation forecast, whereas the ECMWF model is the most accurate. The evaluation results of precipitation detection capabilities under different precipitation thresholds show that ECMWF is more advantageous, especially when the threshold is 25 mm·d^-1. The spatial accuracy test results show that each model performs better in the range of 80°-100°E and 35°-45°N, mainly in central Xinjiang and at the junction of Xinjiang, Gansu, and Qinghai provinces. Out of all the models, the ECMWF model demonstrated the best performance and the KMA model the worst.