Swarm intelligence algorithms (SIAs) have been widely applied to diverse optimization problems, yet they may suffer from high computational cost when confronted with large-scale multi-objective optimization problems (LSMOPs). In this paper, we propose a swarm optimizer with contribution-prioritized dimensionality reduction and surrogate modeling (SO-CPDR-SM). In SO-CPDR-SM, the population is first partitioned into multiple ranks by non-dominated sorting. For individuals of each rank, dimensional contribution analysis is performed, enabling the identification of compact and informative feature subsets tailored to different areas of the search space. To efficiently allocate computational resources, surrogate models are constructed for each rank based on the selected features, facilitating precise approximation of objective functions. Experimental results on complex benchmark problems validate the superiority of SO-CPDR-SM over some state-of-the-art large-scale multi-objective optimization algorithms (LSMOAs).