An investigation was conducted to comparatively evaluate two models predicting the level of crude protein (CP) in soybean meal using Near Infrared Reflectance Spectroscopy technique (NIRS). This study was conducted at the laboratory of quality control in a commercial poultry farm in Peru. The study included 590 data of CP estimated from 295 samples of soybean meal. The data were sorted and classified into two groups, based on the prediction model used, which were the NUTRECO Model (N) and the ADISSEO Model (A). Statistical analyses of variance (ANOVA), means comparison (Tukey), T-Student, correlation, regression, and descriptive statistics, using the GLM procedure of SAS were conducted. It was evaluated the CP values estimated by both models. The results show that the CP values ?? of soybean meal estimated by the models N and A differed (P < 0.05). Model A presented a lower coefficient of variation, lower dispersion and standard error that model N. It is concluded that ADISSEO model estimates the CP content in soybean meal with greater precision than NUTRECO model.