Drones have been widely utilized in logistics delivery, offering advantages of fast delivery and low cost. In this study, we propose a multi-mode vehicle-drone joint delivery model and its path optimization problem to address the needs of customers located far from the warehouse center and facing transportation limitations. This model integrates the vehicle-drone collaborative delivery and parallel delivery models, incorporating trucks equipped with drones and independent drone fleets. Building upon this model, we establish a mixed integer programming model with the objective of minimizing delivery time and design a multi-operator genetic algorithm based on knowledge learning strategies to improve search efficiency. Experimental results demonstrate that the vehicle-drone joint delivery model significantly reduces delivery time compared to traditional delivery methods. The improved genetic algorithm exhibits superior performance on large-scale datasets. The findings of this research provide guidance and reference for tackling the complex and dynamic “last-mile” problem in logistics delivery.