Understanding the relationship between climate dynamics and crop productivity is essential for strengthening agricultural resilience to climate change. This study evaluates the accuracy of ten general circulation models from CMIP6 in modeling temperature and precipitation across three arid sites in Iran from 1990 to 2014. The most reliable models were then downscaled to project climate change under SSP2-4.5 and SSP5-8.5 scenarios for the period from 2026 to 2100. Using the DSSAT CERES, CROPSIM, and NWheat models, wheat yields were projected, with a focus on the effects of adjusted planting dates. Seven GCMs including ACCES-CM2, CMCC-ESM2, EC-Earth3-CC, CNRM-ESM2-1, INM-CM4-8, INM-CM5.0, and NorESM2-MM were identified as the most effective in replicating observed temperature and precipitation patterns. The results showed that substantial increases in annual temperature and reductions in precipitation under both scenarios (SSP2-4.5 and SSP5-8.5) compared to the baseline period. Wheat yield projections across three arid study sites under future climate scenarios (SSP2-4.5 and SSP5-8.5) demonstrate regional variations. In Iranshahr, yields are projected to increase by 16.7% under SSP2-4.5 during 2026-2050, but decline by the end of the century, especially under high-emission scenarios. In contrast, Zabol shows consistent yield reductions, reaching up to 26.32% by the century's end. CO2 fertilization slightly mitigates these losses, with Iranshahr benefiting most (up to 17.69%). Adjusting planting dates further improves yields, particularly in Iranshahr, highlighting the importance of combining strategies for sustaining wheat production in arid regions. However, extreme increases in temperature may limit the effectiveness of these adaptations.