The objectives of this study were to use a multibreed animal model to verify the presence of heterogeneity of genetic variance between Angus and Nelore breeds, to verify the importance of the variance due to the segregation between these two breeds and to estimate breed additive genetic effects on the dominance and additive × additive epistatic effects. It was used in this study 4,016 records of post weaning weight gain from Brangus breed foundation herd from Embrapa Pecuária Sul, analyzed by three different models: multibreed animal model with segregation, multibreed animal model without segregation and traditional animal model. Heterosis was substantially reduced in advanced crosses. The posterior mean genetic variance of the Angus breed (189 ± 26 kg 2 ) was more than twofold of that obtained for the Nelore breed (82 ± 23 kg 2 ) and segregation variance between these two breeds was 11 ± 5 kg 2 . The change in the breed genetic variance estimates was not significant when the segregation variance was set to zero. Bayesian choice criteria point to the multibreed animal model without segregation between breeds as an adequate parsimonious alternative for genetic evaluation of this population.