A neurofuzzy logic controller with a compensating neural network and a fine-tuning mechanism in the consequent membership functions is proposed to design the model-following control of MIMO nonlinear systems. The control strategy is developed to facilitate interconnection compensation among subsystems by the compensating neural network and to realise feedback linearisation by online function approximation. By tailoring the fine-tuning mechanism to overcome the equivalent uncertainty appearing within subsystems or as a result of plant uncertainty, function approximation error, external disturbances, or measurement noise, the system is robust to some extent. The overall neurofuzzy control system is proved to be uniform ultimate bounded by using Lyapunov stability theory. Simulation results of a two-link manipulator demonstrate the effectiveness and robustness of the proposed controller.