This study assessed two versions of the crop model CropSyst (i.e. EMS, existing; MMS, modified) for their ability to simulate maize (Zea mays L.) yield in South Africa. MMS algorithms explicitly account for the impact of extreme weather events (droughts, heat waves, cold shocks, frost) on leaf development and yield formation. The case study of this research was at an experimental station near Johannesburg where both versions of the model were calibrated and validated by using field data collected from 2004 to 2008. The comparison of EMS and MMS showed considerable difference between the two model versions during extreme drought and heat events. MMS improved grain-yield prediction by ∼30% compared with EMS, demonstrating a better ability to capture the behaviour of stressed crops under a range of conditions. MMS also showed a greater variability in response when both versions were forced with scenarios of projected climate change, with increased severity of drought and increased temperature conditions at the horizons 2030 and 2050, which could drive decreased maize yield. Yield was even lower with MMS (8 v. 11 t ha^–1 for EMS) at the horizon 2050, relative to the baseline scenario (∼13 t ha^–1 at the horizon 2000). Modelling solutions accounting for the impact of extreme weather events can be seen as a promising tool for supporting agricultural management strategies and policy decisions in South Africa and globally.