This paper focuses on the modeling and estimation of tail loss distribution parameters from Egyptian’s commercial fire loss severities. Using theoretical extreme value, we use the generalized distribution of Pareto (GPD) and compare it to standard parametric modeling based on exp, Weibull, gumbel, frechet, lognormal and gamma distributions. The goodness-of-fit tests included Kolmogorov-Smirnov, Anderson and Cramer-von Mises test is carried out, and the calculation of the value-at-risk and expected shortfall are performed. We use the bootstrap approach to create confidence intervals for the estimates.