This work evaluates the performance of constraint programming (CP) and integer programming (IP) formulations to solve the Simple Assembly Line Balancing Problem (SALBP) exactly. Traditionally, its exact solution by CP or IP and standard software has been considered to be inefficient to real-world instances. However, nowadays this is becoming more realistic thanks to recent improvements both in hardware and software power. In this context, analyzing the best way to model and to solve SALBP is acquiring relevance. The aim of this paper is to identify the best way to model SALBP-1 (minimizing the number of stations, for a given cycle time) and SALBP-2 (minimizing the cycle time, for a given number of stations). In order to do so, a wide computational experiment is carried out to analyze the performance of one CP and three IP formulations to solve each problem. The results reveal which of the alternative models and solution techniques is the most efficient to solve SALBP-1 and SALBP-2, respectively.