Tail dependence of crude oil price returns between four major benchmark markets are analyzed through the lenses of nonparametric copula models. This paper illustrates that nonparametric copula is flexible to incorporate important empirical patterns of tail dependence and provides better goodness-of-fit to the data than the optimal parametric copula. Estimation results show that the level and the structure of tail dependence of crude oil returns vary significantly depending on data frequency and the time period covered.