3D model retrieval is a research focus at home and abroad. In this paper, we propose a novel 3D object retrieval system via group sparse coding based on multimodel dataset. First, we extract SIFT feature from a series of 2D model images which recorded from each 3D model. Then the visual topic distribution generated by LDA(latent dirichlet allocation) is selected to represent each 3D model. Finally, the sparse coding algorithm is utilized to compute the similarity between different 3D models as to solve the retrieval problem. Experimental results demonstrate the effectiveness of the proposed algorithm.