A variable-grain strategy is proposed to tackle the complex problem of optimization design of fine-pitch gear with multiple objectives and restrictions involved. Multi-objective optimization mathematical models of various grains are established, in order to ensure the tooth profile optimal and reduce the noise generated by the slide of tooth flank meshing and the mutation of friction torque as criteria to evaluate different schemes. The coarse-grained model is utilized in the early design to efficiently determine the direction to the globally optimal solution. Then more accurate model can be utilized and finally the optimal scheme can be acquired on the basis of the fine-grained model. Based on the theory of meshing noise and the superiority criteria, the algorithm is put forward for the fine-pitch gear design problem, and the Pareto optimal sets are obtained.