This paper compares the performance of the different models used to estimate portfolio value-atrisk (VaR) in the BRICS economies. Portfolio VaR is estimated with three different multivariate risk models, namely the constant conditional correlation (CCC), the dynamic conditional correlation (DCC) and asymmetric DCC (ADCC) GARCH models. Risk performance measures such as the average deviations, quadratic probability function score and the root mean square error are used to back-test the performance of the models at 90%. The results indicate that portfolios with more weight to currency and less to equities prove to be the best way of minimizing loses in BRICS.