The article describes a simulation—optimisation system for the management of Eucalyptus grandis stands in Zimbabwe. The simulation sub-system uses an individual tree diameter growth model, a static height model, a stem taper model and mortality models. Diameter growth and height prediction of trees are driven by the development of stand dominant height. Mortality of trees is predicted with a combined use of a stand-level model for the self-thinning limit and a tree level survivor function. All the models have been developed for seedling stands, which means that the simulator cannot be used in coppice forestry. The optimisation sub-system finds optimal values of decision variables (DV) for each thinning and final felling. The DV are stand basal areas, which activate thinning and clear-felling, and the harvest percent of each thinning. The sub-system can use either soil expectation value (SEV), mean annual increment (MAI) or forest rent (FR) as the objective variable. The simulation sub-system was combined with the direct search optimisation algorithm of Hooke and Jeeves. The combined system was used to find optimal management schedules for E. grandis seedling stands on three site classes. When SEV was maximised, the optimum management schedule for stands planted at 1199–3660 trees per ha on site index (SI) 14 (dominant height of 14 m at 7 years) did not include thinning. On SIs 19 and 26, seedling stands planted at 2095 and 3660 trees per ha required a thinning at 6.4 and 10.3 years, respectively, and rotation lengths of 10.5–13.4 years. On medium fertility sites, the optimal management schedules were sensitive to discounting rate and management objective, but insensitive to timber prices and stand establishment costs.