The station instantaneous workload bottleneck was caused by task processing time differences among different models. According to the given production schedule and considering the average workload and instantaneous load of assembly line, mixed assembly line balancing model was built with the objective of minimum station processing time variation, station’s workload smooth index and assembly line overload time. The effect of task processing time differences among different models on the assembly line balance was also taken in to account. The multi-objective optimization algorithm based on non-dominated sorting particle swarm optimization (NSPSO) was designed. The example verified that, compared with genetic algorithm, NSPSO had higher solution quality and solving efficiency in large mixed balancing problem. 