A wide gap still exists between achievable and acceptable grain yield under smallholder farmer’s conditions. Thus, following a participatory rural appraisal, it was suggested that the problem could be overcome by deploying cultivars that combine high grain yield stability and inherent resistance to the main biotic and abiotic stresses. The objective of the current study was therefore to evaluate germplasm that can be used to develop appropriate cultivars for resource-poor farmers in marginal areas. Thus, 24 inbreds that are adapted to southern African environments were crossed in sets, according to a North Carolina Design II mating scheme. The resultant 72 F1 hybrids were evaluated for grain yield under ‘optimum’ conditions and under stress from low moisture and disease at sites representing lowland tropical dry (winter site), mist belt, moist, and dry mid-altitude environments. Preliminary results show significant differences in grain yield among genotypes in all the environments, suggesting that appropriate cultivars can be developed using local germplasm. A significant interaction of gray leaf spot (GLS) and phaeosphaeria leaf spot (PLS) disease was observed at the hotspot site (Cedara) in the mist belt of South Africa. However, there was no direct relationship between leaf blighting due to either GLS or PLS and grain yield. Although more than 80% leaf blighting occurred in the most susceptible check, grain yield was relatively high and ranged from 6 to 12 t ha–1 at the hotspot. This was comparable to the yield obtained in the absence of disease pressure at the high-input site (Rattray Arnold), representing the mid-altitude moist environment where yield ranged from 5 to 13 t ha–1. At the low-input site (Kadoma), which represents the mid-altitude dry environment, grain yield ranged from 4 to 9 t ha–1. Nonetheless, yield was significantly lower in the moisture stress trial that was conducted in winter at Save Valley Experiment Station (lowland tropical dry). Grain yield ranged from 0.8 to 4 t ha–1 under this artificially managed site with low moisture stress at the grain-filling stage. Grain yield was less than 50% of that obtained under optimal moisture conditions at the same site. In the full-irrigation trials representing optimum moisture conditions, the yield ranged from 5 to 8 t ha–1 at low planting density, while higher yield ranging from 5 to 11 t ha–1 was obtained at high planting density. The grain yield data from three summer sites (Rattray Arnold, Kadoma and Cedara) was pooled and subjected to stability analysis. Results showed at least four hybrids that combined high grain yield and stability across the four environments. Preliminary genetic analysis indicates that both additive and non-additive gene action can explain yield under both stress and non-stress conditions, suggesting that both selection and hybridization can be employed to enhance the grain yield.