The Stewart platform has characteristics of low natural frequency, antagonistic design objectives and always no analytical expression of both objective functions and constraints which make it difficult to obtain optimum performance with high dynamic performance. Compared with the traditional adopting indices such as manipulability, stiffness, dexterity, accuracy, etc., lowest natural frequency in the total workspace and ratio of maximum and minimum frequencies at typical configurations are developed as dynamic stability and isotropy indices respectively to improve the dynamic performance of Stewart platform. Evolutionary Multi-objective Optimization (EMO) was introduced synthetically considering the dynamic stability index, isotropy index and global dimensionally homogeneous Jacobian matrix condition number. Two typical EMO algorithms are applied for solving the Multi-objective Optimization Problem (MOP), and the algorithms performance was compared. The results are valuable in designing a Stewart platform for well dynamic capability.