Abstract：Glowworm swarm optimization (GSO) is a newly appeared method for swarm intelligence optimization. Because the GSO algorithm has low precision defects, easily falling into local optimum value and slow convergence speed when solving the optimal value of complex functions, an improved change the step of GSO algorithm was proposed. Step in the algorithm, with the increase in the number of iterations and the curve of diminishing. So at the beginning of iteration the step is bigger and the group due to stronger global search capability, and in later iterations for smaller step that group has a strong local searching ability. With the experimental results on 6 standard test functions, the results show that this method is superior to GSO in simple operation, computational precision and convergence rate.