Particle swarm optimization (PSO) algorithm has good global optimization ability . However, PSO has some disadvantages such as the tendency to easily fall into local extremes, slow convergence speed, and low convergence accuracy at the late stage of the algorithm. Therefore, this study optimizes the traditional PSO algorithm, affording a simplified version and introducing random disturbances to facilitate the falling out of local extremes, thus enhancing its performance on global optimization. Moreover, a hybrid algorithm for the inverse design of photonic devices is proposed by combining PSO and the greedy algorithm with the gradient descent method to evaluate automatic switching between algorithms. Compared with the traditional PSO algorithm, the proposed hybrid algorithm shows better performance on global optimization with a faster convergence speed, higher accuracy, and superior design efficiency. A 1∶1 optical splitter is inversely designed using the proposed hybrid algorithm. At a bandwidth of 120 nm, the range of insertion loss at the output of the device is 0.125 dB?0.197 dB. Moreover, the device is manufacturable robustness.