The aim of this study was to develop a generalized height-diameter model for predicting tree height of Acacia mangium plantations in South Sumatra that could account for the variability of site and stand conditions. Six commonly used non-linear growth functions (i.e. Gompertz, Chapman-Richards, Lundqvist-Korf, Weibull, modified logistic, and exponential) were selected as candidate base models and were fitted to individual tree's height-diameter data of A.mangium plantations. A total of 13,302 trees collected from permanent sample plots with various spacing , stand age, and site quality were available for this study. The data were split into two sets: one set being the majority (75%) was used to estimate model parameters and the remaining data set (25%) was used to validate the models. The results showed that the six base models produced almost identical fits with a relatively high root mean squared error (± 3.4 m) and a relatively low proportion of the total variation in obser ved tree height (52.5 - 53.4%). The Lundqvist-Korf (LK) model performed slightly better than the other models based on the goodness of fit as well as bias and standard errors of the predictions. This LK model can be fitted easily and provided more satisfactory fit when additional variables were included into the model, hence was selected as the base model. Introducing stand variables into the selected base model resulted in a significant improvement of the accuracy for predicting heights. The root mean squared error decreased by the value between 0.5564 and 1.4252 m and the proportion of variation explained by the model increased by the value between 13.88 and 33.21%. The best improvement based on fit and model validation was achieved by the generalized height-diameter model with inclusion of stand age and site index.