[Introduction]: A new site classification system for Tectona grandis L.F. is presented based on information collected during more than 30 years of research, in order to contribute to the silvicultural management of one of the most important species in Costa Rica. [Objective]: Build updated site index curves for teak plantations. [Methodology]: The data come from the fustal analysis of 55 of dominant trees aged between 6 and 40 years, at cutting age, selected in different locations of the Pacific Vertiente. These were combined with dominant height measurements from 38 permanent plots measured between 10 and 32 years that were part of 6 formal growth and yield trials. Growth models were defined, adjusted for non-linear regression of mixed effects, considering that these are repeated measurements annually. We worked with six mathematical models: Richards, Gompertz, Logistics, Korf, Mitscherlich and Weibull. Prior to adjustment, the models were reparameterised according to the observed dominant height (S), at the base age of 16 years, associating it to a random effect. The best model was chosen according to information criteria and good fit statistics. [Results]: The information criteria of Akaike and Bayesian indicate that Korf’s model was superior to the others, both in its anamorphic and polymorphic variants; the goodness of adjustment statistics confirmed this assessment. The analysis of the residues showed a better fit for this model by projecting growth at dominant height with age. Although the polymorphic model was superior, the observed trend seems to rule out the existence of a common asymptote for teak plantations growing in different site qualities. [Conclusions]: The results showed that fustal analysis and permanent plot data present similar patterns, which are adequately represented by the Korf model.