Anion–π interaction is a new type of non-covalent interaction. It has attracted growing interest in recent years both theoretically and experimentally. However, the nature of bonding between an anion and an electron-deficient aromatic system has remained elusive. To understand the bonding nature in depth, we have carried out a systematic computational study, using model systems that involve tetraoxacalix[2]arene[2]triazine 1, an electron-deficient macrocyclic host, and four anions, X⁻ (X⁻ = SCN⁻, NO₃⁻, BF₄⁻, and PF₆⁻), of varied sizes and shapes. The geometries for the 1·X⁻ complexes were optimized using the extended ONIOM (XO) method. The good agreements with the X-ray experimental results provide a validation of our theoretical schemes. The nature of the non-covalent interactions was analyzed with the help of the AIM (atoms in molecules), RDG (reduced density gradient) and LMO-EDA (local molecular orbital-energy decomposition analysis) methods. The results clearly reveal the involvement of anion–π bonding, as well as a weak, yet significant, hydrogen bonding interaction between the benzene C–H on 1 and the anion of NO₃⁻ or PF₆⁻. The bonding energies of 1·X⁻ were calculated with the XYG3 functional, and the results were compared with those from MP2, M06-2X and some other functionals with non-covalent interaction corrections (e.g., B3LYP-D3, and ωB97X-D). We conclude that the binding strengths follow the order of 1·NO₃⁻ > 1·SCN⁻ > 1·BF₄⁻ > 1·PF₆⁻, where the difference between 1·SCN⁻ and 1·BF₄⁻ is less significant. The strongest interaction in 1·NO₃⁻ comes from: (1) the effective electronic interaction between NO₃⁻ and the triazine rings on 1; and (2) the weak hydrogen bonding interaction between the benzene C–H on 1 and nitrate, which cooperates with the anion–π interactions.