This paper is concerned with the identification and nonlinear predictive control approach for a nonlinear process based on a third-order reduced complexity, discrete-time Volterra model called the third-order S-PARAFAC Volterra model. The proposed model is given using the PARAFAC tensor decomposition that provides a parametric reduction compared with the conventional Volterra model. In addition, the symmetry property of the Volterra kernels allows us to further reduce the complexity of the model. These properties allow synthesizing a nonlinear model-based predictive control (NMBPC). Then we construct the general form of a new predictor and we propose an optimization algorithm formulated as a quadratic programming (QP) algorithm under linear and nonlinear constraints. The performance of the proposed third-order S-PARAFAC Volterra model and the developed NMBPC algorithm are illustrated on a numerical simulation and validated on a benchmark such as a continuous stirred-tank reactor system.