In this work, we proposed a revised ant colony optimization (ACO) algorithm to achieve computationally efficient long horizon model predictive control (LHMPC). LHMPC has the benefit of better steady-state control performance in comparison with short-horizon predictive control solutions. However, the computational burden of LHMPC increases exponentially with the number of the prediction horizon, which imposes a large challenge for the real-time implementation. ACO emulates the behavior of real ants to explore the shortest route between their nest and foods. It is a widely recognized solution for different combinatorial optimization problems and has the potential to efficiently solve LHMPC optimizations. A thorough comparison between the proposed and the well-known branch and bound (BnB) model predictive control technique is carried out on a two-level power converter-fed synchronous reluctance motor (SynRM) system. Simulation data reveals that the transient and steady-state performances of both methods are similar. However, the search efficiency in terms of the number of visited nodes per sampling instant of the proposed ACO LHMPC solution achieves a significant improvement (around 35.9% faster).