Aiming at the disadvantages of the salp swarm algorithm (SSA), such as slow convergence speed and easy to fall into local optima when solving complex problems, an improved SSA equipped with gravitational search technique and normal cloud generators(CGSSA) is proposed. The acceleration coefficient of the gravitational search algorithm (GSA) is introduced in the stage of updating the position of the leader of salps, which avoids the invalid search of the salp swarm and accelerates the search speed. The normal cloud model is used to update the position of the followers of the salp, which enriches the diversity of the population. At the same time, the entropy value of the normal cloud model can be adaptively adjusted with the increase of iteration times, which effectively improves the convergence accuracy in the later iteration period. A comprehensive comparison between the CGSSA and other 10 optimization algorithms is made on 23 benchmark functions. The statistical result, convergence curve and box–whisker plot of simulation experiment show that the improved algorithm has better performance in search efficiency, convergence accuracy and avoiding local optimum.