In this paper, authors consider Bahadur asymptotic efficiency of MLE {\bar β}_{ML} of β, which is an unknown parameter vector in the semiparametric regression model Y_i=X_i~T β + g(T_i) + i, where g is an unknown Holder continuous function, i is a random error with density function (.), X_i is a known vector in R_k, T_i is a random variable in [0, 1]with density function T(t).