||Height–diameter (H–D) models can be used to predict missing tree heights from diameter at breast height measurements for many forest management applications. A major challenge in developing H–D models for tropical forests is many rare tree species with very few observations. This study utilized the taxonomic hierarchy of genus and species as random effects in a non-linear mixed-effects modelling approach to develop species-specific H–D models for 842 species representing 295 genera. Data were collected from eight forest reserves of hill Dipterocarpacae tropical forests in Peninsular Malaysia. Variations of the Wykoff (WYK) and von Bertalanffy-Richards (VBR) models with covariates were evaluated. Results showed that WYK outperformed VBR with the latter having convergence issues and larger root-mean-squared errors than the former. Among the WYK models, weighted WYK with maximum tree height as stand-level covariate (WYK-HMAX-WT) had the best fit, but it underpredicted heights for larger trees. Nonetheless, the mixed-effect modelling strategy behind WYK-HMAX-WT resolved the need to aggregate species during model development by using the strength of the entire dataset to derive parameter estimates, especially for 36 per cent of the 842 species having less than four observations. Lastly, WYK-HMAX-WT has the flexibility of predicting heights on a regional basis, while retaining the capacity to adjust to stand conditions. The outlined approach should work well for other predictive models and/or regions with a high variety of species present.