In this paper, we focus on three daily exchange rate returns dynamics. Indeed, we have assessed five GARCH-class models under three alternative distributions. Our findings confirm that the skewed Student-t FIAPARCH model performs very well. Then, we have computed short and long Value-at-Risk and Expected Shortfall based on AR (1) – FIAPARCH under normal, Student-t and skewed Student-t distributions. More precisely, we have investigated the estimation performance by computing both In-sample and Out-of-sample VaR for one-day-ahead horizon. Results reveal that VaR and ES estimations based on skewed Student-t FIAPARCH models outperform other models for both long and short trading positions.