To improve the efficiency of training the T-S fuzzy model based RBF neural network, the Levenberg-Marquardt algorithm is introduced into it, which speeds up the convergence and reduces the probability for the training to get into the local minimum point. Next, a kind of more efficient algorithm, named hybrid learning algorithm，is proposed. At last, the efficiency and practicability of the Levenberg-Marquardt algorithm for the training of the T-S fuzzy model based RBF neural network are tested through an experiment.