Abstract: Based on response surface method (RSM), a highly efficient approximate method for computing the reliability index or failure probability of slopes is proposed. The Morgenstern-Price method is adopted to compute sample points instead of traditional time-consuming finite element method in the construction of response surface functions, and the computing time is greatly decreased. By making use of the random sampling idea of Monte Carlo technique, a new random search algorithm is presented, which can locate the critical non-circular slip surface associated with the minimum reliability index. Two examples are given to verify the practicability and reliability. It is also found that when separately taking the minimum safety factor and minimum reliability index as the target function their critical slip surfaces show significant difference. In the end, the influence of the probability distribution and correlativity of soil parameters on the reliability index of slope is discussed.