Energies computed by B3LYP and other popular DFT functionals are flawed by systematic errors, which can become considerable for larger mols. These errors, predominately due to inadequacies in assessing longer-range nonbonded attractive effects (dispersion), are illustrated by the isodesmic stabilization energies of n-alkanes (based on methane and ethane, which have no stabilizing 1,3-alkyl group interactions). Newer functionals, designed to describe weak interactions, give somewhat better agreement with expt., but are not fully satisfactory. [on SciFinder (R)]