In order to realize the lightweight of the high-horsepower tractor gear box under the requirements of strength and stiffness, a multi-objective optimization method for high-horsepower tractor gear box based on an improved non-dominated sorting genetic algorithm-Ⅱ (NSGA-Ⅱ) was proposed. Firstly, the force of the gear box was analyzed, and the statics simulation analysis method of the load-bearing gear box was proposed. The strength and stiffness performance of the gear box were simulated and analyzed by ANSYS. Then, K-means clustering algorithm, normal distribution crossover (NDX) and differential mutation search strategy were introduced to improve the NSGA-Ⅱ algorithm, and example tests were carried out. The results showed that the solution set distribution uniformity and stability of the improved NSGA-Ⅱ were better than NSGA-Ⅱ, and the optimization effect was better, which verified the effectiveness and superiority of the proposed algorithm. Finally, the multi-objective optimization of the gear box was carried out based on the improved NSGA-Ⅱ algorithm, and the simulation analysis and verification of the optimized gear box were carried out. The results showed that the mass, maximum deformation and maximum stress of the optimized gear box were 168.16kg, 0.449mm, and 215MPa, respectively, which were better than 168.16kg, 0.454mm, and 216.12MPa of NSGA-Ⅱ. That realized the lightweight of the high-horsepower tractor gear box under the requirements of strength and stiffness, and further verified the effectiveness and superiority of the proposed algorithm in solving the multi-objective optimization problem of high-horsepower tractor gear box, and the research results provided a method reference for the simulation and optimization of the high-horsepower tractor gear box.