This work presents a detailed evaluation of the performance of density functional theory (DFT) for the prediction of zero-field splittings (ZFSs) in high-spin nitrenes. A number of well experimentally characterized triplet mononitrenes, quartet nitrenoradicals, quintet dinitrenes, and septet trinitrenes have been considered. Several DFT-based approaches for the prediction of ZFSs have been compared. It is shown that the unrestricted Kohn-Sham and the Pederson-Khanna approaches are the most successful for the estimation of the direct spin-spin (SS) interaction and the spin-orbit coupling (SOC) parts, respectively, to the final ZFS parameters. The most accurate theoretical predictions (within 10%) are achieved by using the PBE density functional in combination with the DZ, EPR-II, and TZV basis sets. For high-spin nitrenes constituted from light atoms, the contribution of the SOC part to ZFS parameters is quite small (7%-12%). By contrast, for chlorine-substituted septet trinitrenes, the contribution of the SOC part is small only to D value but, in the case of E value, it is as large as the SS part and has opposite sign. Due to this partial cancellation of two different contributions, SS and SOC, the resulting values of E in heavy molecules are almost two times smaller than those predicted by analysis of the widely used semiempirical one-center spin-spin interaction model. The decomposition of D(SS) into n-center (n=1-4) interactions shows that the major contribution to D(SS) results from the one-center spin-spin interactions. This fact indicates that the semiempirical SS interaction model accurately predicts the ZFS parameters for all types of high-spin nitrenes with total spin S=2 and 3, if their molecules are constructed from the first-row atoms.