Pepper (Capsicum sp.) is important for the Brazilian agribusiness, serving as raw material for the food, pharmaceutical and cosmetic industries. The adequate evaluation of its plants growth may help in understanding the causes of crops yield variation, with it being able to be studied by regression models, which help to adequate the management with the different phenological phases. This study aimed to compare the fit of linear Polynomial model and the Logistic and Gompertz nonlinear models in the description of pepper plants growth from the Doce cultivar. Estimates were obtained by the Gauss-Newton method, with the quality of fitted models compared by graphical analysis and evaluators: adjusted coefficient of determination (R 2 adj), Residual Standard Deviation (RSD) and the corrected Akaike Information Criterion (AICc). Residues were normal, independent and homocedastic at 5% level of significance. All models properly described the height of the Doce cultivar. The Logistic model was the most adequate according to the fitting evaluators, having higher value of R 2 adj , and lower RSD and AICc values.