Accurate simulation of climatic parameters such as air relative humidity (RH) is essential for understanding regional climate dynamics, particularly in climatically diverse regions. This research evaluates the performance of seven CMIP6 climate models in simulating RH across two cities with distinct climates in Iran: Rasht (humid) and Yazd (arid). For this purpose, historical RH data (1990-2014) were compared with model outputs using the Kling-Gupta Efficiency (KGE) metric, bias correction via Quantile Mapping (QM), Corrected Ratio of the Mean (CRM) and ensemble modeling. In Rasht, only three models yielded positive KGE values, with CanESM5 performing best (KGE=0.21), while the EM achieved a higher accuracy (KGE=0.29) but slightly underestimated RH (CRM=-0.00176). In Yazd, all the models showed acceptable performance, with CanESM5-CanOE leading (KGE=0.72) and the EM outperforming individual models (KGE=0.77 and CRM=0.00024). Future RH projections (2026-2050) under three SSP scenarios (SSP1-2.6, SSP2-4.5, SSP5-8.5) revealed that Rasht may experience significant RH declines under SSP5-8.5, while the RH trends in Yazd remain relatively stable across these scenarios. These findings highlight the importance of ensemble modeling and bias correction in improving RH simulations, particularly in arid regions.