In order to solve the problems of the large sorting quantity of cigarette in tobacco logistics distribution center and the great impact of assignment of cigarette specification on the total processing time of orders, the work aims to study the allocation of each sorting zone and improve the sorting efficiency. A mathematical model with the objective function of minimizing the similarity coefficients of specification in each zone was developed and solved by an improved genetic particle swarm dynamic clustering (GAPSO-K) algorithm. Firstly, the similarity coefficient of each specification was improved by combining the sorting quantity of each specification as the fitness function. Then, the inertia weight factor was improved in the particle swarm algorithm so that its value could be changed adaptively. Finally, the cross-variance in the genetic algorithm was introduced in the particle swarm dynamic clustering algorithm to expand the search range of the solution, and the results were compared with the other algorithms based on Matlab. The results were simulated and verified in EM-plant. Combined with the data simulation verification in a tobacco logistics distribution center, the time for processing order with GAPSO-K algorithm was 234.5 s, which was significantly reduced compared with the traditional time, effectively improving the efficiency of flexible logistics sorting. The use of this algorithm can give full play to the goodness of both algorithms, with better convergence and merit-seeking, and provides a new idea for flexible logistics product rule allocation.