Logistics distribution plays a significant role in today’s rapid development of urbanization, and efficient vehicle route planning in the distribution process has a significant impact on improving distribution efficiency and reducing logistics costs. Traditional methods have certain limitations when dealing with problems of different scales. To minimize the total transportation cost of the Capacity Vehicle Routing Problem (CVRP), we incorporate Particle Swarm Optimization (PSO) and Graph Attention Network (GAT) into the Ant Colony Optimization (ACO) framework. By integrating PSO to dynamically adjust the pheromone update strategy of ACO and leveraging GAT’s adaptive feature extraction capability, a three-layer collaborative optimization framework is established to enhance both global exploration and local exploitation. Finally, the performance of the proposed algorithm is compared with other algorithms using five test datasets of different sizes. Three sets of ablation experiments are conducted to demonstrate the effectiveness of the algorithmic components. The experimental results show that the proposed method achieves significant improvements in solving the capacity-constrained vehicle routing problem.