This study was conducted to evaluate the performance of CMIP6 climate models and their ensemble output in the statistical downscaling of precipitation over Hormozgan Province using the Empirical Quantile Mapping (EQM) method. For this purpose, monthly precipitation data from four coupled atmosphere–ocean general circulation models, namely BCC-CSM2-MR, MPI-ESM1-2-HR, CESM2-WACCM, and GFDL-ESM4, for the period 1985–2014, together with observations from seven synoptic stations (Minab, Bandar Abbas, Bandar Lengeh, Siri, Jask, Kish, and Abu Musa) located in Hormozgan Province, were used. Model performance was evaluated using the Kling–Gupta Efficiency (KGE) index and Taylor diagrams. To reduce uncertainty, an ensemble model was developed using a rank-based weighted averaging approach. All computations were performed in the R software environment. The findings showed that the EQM method performed effectively in correcting precipitation bias in the climate model outputs. Among the individual models, GFDL-ESM4 showed the best performance, with KGE values ranging from −0.05 to 0.14, whereas MPI-ESM1-2-HR exhibited the weakest performance across all stations, with KGE values ranging from −0.22 to −0.66. The highest accuracy of GFDL-ESM4 was recorded in the central and eastern parts of the province, particularly at Minab (KGE = 0.13) and Bandar Abbas (KGE = 0.14). The results also indicated that the ensemble model, with KGE values ranging from −0.16 to 0.01, did not outperform the best individual model (GFDL-ESM4) in simulating precipitation over the study area. In most stations, GFDL-ESM4 provided more accurate results than the ensemble model, for example at Minab, where its KGE value was 0.22 higher than that of the ensemble.