The objective of this study has been to develop a prediction equation of fat-free mass (FFM) from buffalo calves. Twenty buffaloes were fed ad libitum access at unifeed, with vitamin-mineral integration, for 14 months. Seven days before slaughtering, the animals were weighed and bioelectrical impedance measurements were collected. The data were analyzed by multiple linear regressions to evaluate the relationship between FFM and various predictor variables. Stepwise regression was used to eliminate variables that did not influence variation in the model. The value of resistance collected showed a decrease when the electrical frequency increases, while the values of reactance (Xc) increase. When using live weight (LW) and reactance at 500 and at 1000 kHz as independent variables, we obtained the best R2 Adj (0.967) and Durbin Watson statistic (2.596) that explain the prediction model (FFM = - 30.59 + 0.993LW + 0.150Xc500 - 0.123Xc1000 + 9.11). These results indicate that the use of bioelectrical impedance analysis has excellent potential as a rapid method, with minimal perturbation for the animal, to predict FFM in buffalo.