Pasture height measurement and its statistical distribution have been neglected topics in tropical pasture research. Thus, the objective of this study was to fit the single normal (SN), double normal (DN), log normal (LN), gamma (GM) and Weibull (WB) models to relative frequency distributions of pasture height (PHT, cm) of native gramma pastures of the Mexican humid tropics. PHT measurements were done in a study where fixed rotational grazing was used (3-day grazing/27-day rest paddock–1) at three stocking rates of two, three and four F1 (Holstein × Zebu) cows ha−1. Two paddocks per stocking rate were used. Approximately 400 PHT measurements were done on each paddock, before (22 times) and after (23 times) grazing. The models were fit by least squares, using the Marquardt–Lavenberg iteration. The Akaike’s information criterion corrected (AICc) was used to select the model with the best fit. Before grazing, the models with the best fit were the DN and the GM, and after grazing, the best fit was shown by the DN and the LN. With respect to stocking rate, the best models were: LN and GM for two cows ha−1; DN and LN for three cows ha−1; and DN for four cows ha−1. However, the differences between the AICc values of the model with the best fit and the second in rank were small; thus, either could fit the data equally well. In conclusion, PHT of native gramma pastures did not distribute normally, and either the double normal or the log normal models can be used as means to describe the PHT frequency distributions of this type of pastures.