Abstract: The performance of the model forecast of the June 26 heavy rainstorm in Mianning was tested in detail by using meteorological observation data and multi-models forecast data. The results show that: (1) For the 24h cumulative precipitation forecast, meso-scale regional model had obvious advantages, and the precipitation level and fall zone were the best. The distribution of fall zone of GRAPES_3KM was highly consistent with the actual situation, and it’s TS score of large precipitation over 50mm was the highest. (2) The maximum hourly rainfall intensity of GRAPES_3KM over 10mm matched the actual precipitation fall zone best, followed by the revised product of accumulated precipitation of multiple physical quantities in ECMWF 24h and the probability product of short-time heavy precipitation. (3) The 24h cumulative precipitation extreme points of SWC and GRAPES_3KM models were slightly norther than the actual situation, and the magnitude was smaller. For the peak time of hourly precipitation, SWC model was 4h earlier and GRAPES_3KM was 3h earlier. (4) The forecast of GRAPES_GFS situation field was closer to the actual situation, and SWC model could better predict the existence of mesoscale convergence system over Mianning.