Abstract: With the continuous advancement of numerical forecasting techniques, the meteorological community has imposed higher requirements on the evaluation of high-resolution precipitation forecasts. Conventional precipitation verification approaches can only provide a comprehensive assessment of forecast skills, which fail to capture the intrinsic distribution of forecast errors and identify the error sources. To gain deeper insights into error characteristics of extreme precipitation forecasts in high-resolution models, a case study of a mixed-type heavy precipitation event in North China in 2024 was selected for comprehensive analysis. A verification methodology based on scale analysis was employed to evaluate forecast errors in high-resolution precipitation predictions, and the consistency and reliability of the conclusions were validated using verification results from the summer of 2023. The case study results indicate that the verification results exhibit a high degree of consistency across various models of different resolutions. Notably, the models exhibit the highest forecasting skill score in predicting moderate rain, whereas forecast errors for heavy and torrential rain are predominantly concentrated at a scale of 48 km, with the mean squared error percentage (M_SE%) exceeding 20%. For extremely heavy rain, the models exhibit minimal forecasting skill at the typical storm scale. The seasonal verification of the CMA-MESO regional model further indicates that forecast errors for heavy and torrential rain are concentrated within the 24–96 km scale range, confirming the reliability of the verification results. This scale range has been identified as the critical error scale, primarily caused by the 24-hour time integration of initial errors in small-scale moist convective processes. Physical processes related to convective perturbation weather processes associated with this scale, such as gravity waves and the triggering of baroclinic instability, warrant particular attention.