The problem of optimal density over time for even-aged, mixed-species stands is formulated as a nonlinear-integer programming problem with numbers of trees cut by species and diameter class as decision variables. The model is formulated using a stand-table projection growth model to predict mixed-species growth and stand-structure. Optimal thinning and final harvest age are estimated simultaneously using heuristic random search algorithms. For sample problems with two species, random search methods provide near-optimal cutting strategies with very little computer time or memory. Optimal solutions are estimated for problems with eight initial species/diameter class groups, projected for up to three discrete growth periods. Such solution methods merit further study for evaluating complex stand- and forest-level decisions. Forest Sci. 31:303-315.