The work aims to propose a variable-step cuckoo search algorithm (denoted as LFCS) based on Levi's flight detection probability to solve the problems in solving complex high-dimensional function optimization, such as insufficient accuracy and easy to fall into local optimization. Under the same environment, six test functions of different difficulty and different types were selected, and the LFCS algorithm was compared with the IPSO, IDE, IABC, and CS algorithms to analyze the convergence speed and convergence accuracy of the algorithm. Compared with the other four algorithms, the LFCS algorithm had fewer iterations, faster convergence speed, and higher convergence accuracy. No matter whether it is a low-dimensional function or a high-dimensional function, the LFCS algorithm has improved convergence speed and convergence accuracy. Especially for complex high-dimensional function optimization problems, the LFCS algorithm can find the optimal solution faster and more accurately when the value range is large.