Using the precipitation observation data of the National Meteorological Station and based on the model evaluation tools test system, an evaluation was conducted on the precipitation prediction performance of the European Centre for Medium-Range Weather Forecasts (ECMWF), Global Assimilation Prediction System of China Meteorological Administration (CMA) Global Forecast System, Mesoscale Weather Numerical Prediction System of CMA CMA-SH9 during the 2022 Gansu main flood season rainstorm. The Mesoscale (CMA-MESO), and Shanghai Numerical Prediction Model of results demonstrate that: (1) The four numerical models show similar accuracy in predicting sunny and rainy weather, all above 0.80, with CMA-SH9 achieving the highest score for precipitation forecasts ≥50 mm, followed by CMA-MESO. (2) According to the characteristics of the impact system and circulation situation, the rainstorm in the main flood season in Gansu Province can be divided into three types: subtropical high marginal, northwest airflow, and low trough. The CMA-SH9 model has the best prediction effect for precipitation ≥50 mm, particularly for low-trough rainstorm processes, followed by the subtropical high marginal type. (3) Method for Object-based Diagonostic Evaluation spatial test indicates that the CMA-MESO and CMA-SH9 regional models outperform the ECMWF and CMA-GFS global models in forecasting ≥50 mm precipitation. In the rainstorm process on July 14, the CMA-SH9 model predicted that the centroid distance and axial angle deviation of the rainstorm were relatively small, and the prediction of rainstorm location and spatial trend was most close to the actual situation.