Social networks represent real-world social structures and are essential for linking people across the globe. Determining key nodes in social networks is a challenging issue in the realm of network science. For solving the key node identification issue, numerous centrality measures have been developed, including degree, eigenvector, gravity index, etc. However, the gravity models relied on universal gravity law are unique measures that have grasped the attention of numerous researchers. In this study, we analyze eight previous and recently developed gravity models to identify the best gravity model with respect to three parameters: accuracy, distinguishing ability, and efficiency. Hence, distinct experiments are conducted on social networks utilizing the eight gravity models. The Kendall tau, SIR model, Jaccard similarity coefficient, and ranking monotonicity are utilized as performance metrics for assessing the efficacy of eight gravity models. The empirical findings indicate that the H-index-based gravity centrality has superior accuracy and distinguishing ability in contrast to other gravity models. Finally, we outline our plans for future gravity model-based research.