In order to solve the problem of high transportation costs in the process of multi-vehicle loading, a multi vehicle loading problem that considers both transportation costs and transportation safety was studied. Firstly, under the premise of infinite vehicle loading capacity, a mathematical model for vehicle loading was constructed to minimize the cost of using vehicles and consider load balancing for the problem of vehicle loading of multiple vehicle types and varieties of goods. Secondly, variable neighborhood search algorithm was proposed to solve the problem. Finally, based on the idea of "personification", neighborhood operators such as tail box operator, large vehicle to small vehicle operator, one large vehicle to two small vehicle operator, two small vehicle to one large vehicle operator, and shaking operator to heavy-duty vehicles were designed and applied to the variable neighborhood search algorithm. The example data results demonstrate that compared with traditional intelligent optimization algorithms, the variable neighborhood search algorithm constructed in this paper has strong global search ability. The model and algorithm can reduce the number of vehicles from 4 to 3 while ensuring load balance, thereby reducing vehicle costs by 17.44%. The average loading rate is increased from 0.69 to 0.89, and improving the average vehicle loading rate by 28.99%. The optimization effect is significant. The model and algorithm constructed in this article can greatly save the transportation costs of logistics enterprises, improve transportation profits, and help logistics enterprises improve quality, reduce cost, and increase efficiency while ensuring transportation safety.