Abstract: The forest dominated by Cyclobalanopsis glauca is one of the main types of the subtropical evergreen broad-leaved forests in China. The studies on the primary productivity of the community have been completed with harvesting method. In this paper, a group of optimized regression biomass models for the tree and subtree layer were reported. The models were also analysed, examined and evaluated. The results showed that there is an allometric growth relationship among the organ variables. It is essential to establish various optimized biomass models using the regression analysis method on the basis of various layers, species and variables. The models are relatively of high accuracy and can effectively predict the biomass of the community. The effective combination of the interspecific models are suitable for subtrees and companion trees. The model differences among species are closely related to their states and function in the community. The distribution ratio of net primary productivity in various organs is the synthetical reflection of the community structure and function in interspecific relationship, niche differentiation and population dynamics.