<p>With the advent of computation and the progressive advancement of geotechnologies, the use of mathematical models of wildfire risk became more expressive in quantitative terms. The existence of a significant and variable number of models of wildfires risk presents the necessity to select the model that best fits a probable region – which can be done in two ways: visually or through the confrontation of existing models. In this sense, the present study aims at selecting the wildfire risk models FWI, FMA⁺ and RIF-Database for the Eucalyptus plantations. The study area extends from the north-central coast of the state of Espírito Santo and the south coast of Bahia, Brazil. The database was comprised the period between January 1st, 2010 to June 30st, 2016, with 10,447 occurrences. The validation and choice of the results were determined by the success percentages and the skill score value, for each subzone and risk model. After, it was performed the parametric analysis of Variance Analysis (ANOVA), if the F test was significant, the Tukey-Kramer post-hoc test (p < 0.05) was used to compare all the ids with each other. We observed in days with wildfire, a greater sensitivity of the FMA⁺ model, which presented success percentage values higher than 60 % for all subzones. However, it presented the worst results for the days without wildfire and, consequently, the worst results for the overall success rate. Additionally, considering the skill score value, the FWI model presented the best results for subzone 1. The RIF-Database model presented excellent results, being the model to be used for subzones 2 and 3.</p>