Two adaptive predictive control schemes based on fuzzy models are introduced to control of nonaffine nonlinear systems. The one-step-ahead controller uses a fuzzy model to predict the system output one-step-ahead and minimizes the discrepancy between the system output and the desired trajectory at the next step. In the multi-step-ahead scheme, a group of fuzzy models are employed for a long-range prediction of the system outputs over a prediction horizon, and the cost function involves multi-step tracking errors. To compute the control actions, in both schemes intuitively appealing algorithms are proposed to optimize the cost functions, by making use of the internal special properties of the fuzzy system models. Performances of both schemes are analyzed and an example is given to show the effectiveness and compare the schemes.