A stage‐structured model for projecting stand growth and yield is coupled with a gradient‐based algorithm to determine optimal treatment regimes for Norway spruce (Picea abiesKarst.) plantations. The growth model forecasts the movement of trees between 3 cm diameter classes using a growth matrix that includes nonlinear equations for diameter growth. Because harvest treatments may be specified by tree diameter class, the model allows the determination of the optimal thinning timing, thinning intensity, and thinning type simultaneously with planting density and clearfelling age. The optimization model is used to determine treatment regimes with and without thinning for a wide range of values for economic and biological parameters. Results show that the optimal economic planting density is highly sensitive to interest rate, logging costs, planting costs and site class. For a given set of economic parameters, the optimal planting density is not sensitive to the number of thinnings performed or the thinning intensity. Top thinning is the optimal method for every regime in which thinning is scheduled. Results indicate that managers have great flexibility in choosing thinning schedules: few thinnings and shorter rotations give almost the same economic results as more frequent thinnings and longer rotations. Major uncertainties in these results and areas for further research are discussed.