In agroforestry systems (AFS), where environmental conditions are highly variable at small spatial scales, the use of uniform genetic material of a single cultivar commonly grown in monoculture cropping might not be optimal. However, the use of composite cross populations (CCPs) that contain an inherent genetic variability might be a promising approach under the environmental variability created by trees in AFS. In this experimental trial, the performance of a CCP (‘CC-2k’) of winter wheat was compared to a commercial variety (‘Wiwa’) in a split-plot design at two AFS (Feusisberg and Wollerau) in Central Switzerland. The factor variety (p < 0.05) and the interactions of site and distance to tree (p < 0.05) and site and variety (p < 0.05) significantly affected wheat yield at the plot level. In Feusisberg, CC-2k and Wiwa yield (across all distances) amounted to 1.9 ± 0.7 Mg ha− 1 and 2.0 ± 0.8 Mg ha− 1, respectively, while in Wollerau, to 1.9 ± 0.9 Mg ha− 1 and 0.7 ± 0.4 Mg ha− 1, respectively. Wiwa had a higher protein, Fe and Ca content than CC-2k. Therefore, while the CC-2k outperformed Wiwa in terms of yield in one of the two AFS, Wiwa outperformed CC-2k in terms of quality. Thus, the composite cross population might be better adapted to the heterogenous environment of agroforestry systems but fails to reach the high-quality product of modern cultivars. Further improvements in terms of quality might make CCPs a viable option for diversified agricultural systems with larger environmental heterogeneity than common monoculture cropping systems.