A neural network force approximation model for large atomic systems was suggested, practical use cases were pointed out. This model was trained on ZnO cluster examples calculated using the density functional theory (DFT). Training results were analyzed and compared to DFT results.