An important area of Eucalyptus tereticornis plantations has been established in the Atlantic coast in Colombia. Decision planning and estimates of future wood supply will necessary require the use of growth projection systems. Among the components of this system are models able to predict the total height of trees growing under diverse conditions of stand age, site and management. This research focused on the construction of regional height-diameter models in order to implement a growth simulator. Ten regional height-diameter models were evaluated, which had different numbers and types of predictor variables. Data required were obtained from 63 permanent sample plots established in plantations with stand ages between 2 and 8 years old and stand densities from 388 to 1,640 trees per hectare distributed in the Atlantic coast. Measurements of bias, precision and error were used for model evaluation. During the validation process no important differences among the evaluated models were detected. However, models that contained the mean height and mean diameter of the dominant trees as stand-level predictor variables showed a minor bias and error. Finally, the three-parameter regional model proposed by Krumland and Wensel was selected as the best.