The binding energy is of great importance in understanding the formation and stability of noncovalent interactions. However, the determination of the binding energy with high precision and efficiency in medium- and long-range noncovalent interactions is still challenging for quantum chemistry. Here, we assess the performance of random-phase approximation (RPA), a fully non-local fifth-rung of the Jacob ladder functional, in determining the binding energy of cation-π systems (cation = Li+, Na+, Be2+, Mg2+, Al+, and NH4+; π = C6H6), which, to the best of our knowledge, has not been investigated. Using experimental results as the benchmark, we systematically compared the RPA method to the other ab initio methods (DFT/B3LYP, MP2, CCSD(T), and QCISD(T)) both in calculation accuracy and efficiency. From the perspective of accuracy, RPA is the best among these approaches, followed by the CCSD(T) and QCISD(T) methods. DFT/B3LYP and MP2 provide the worst accuracy. In addition, the computational efficiency of RPA is much faster than that of CCSD(T) and QCISD(T). We believe that RPA is a robust method for the precise description of medium- and long-range noncovalent interactions and is capable of providing benchmarking data. The interaction strength and interaction nature of cation-π systems are further analyzed by atoms in molecules (AIM) and the color-mapped reduced density gradient (RDG) isosurface, which are consistent with the characteristics of a typical cation-π interaction.