A highly diverse and accurate database of 3366 total atomization energies calculated at the CCSD(T)/CBS level by means of W1-F12 theory

The total atomization energy (TAE) is the most challenging thermochemical property for benchmarking density functional theory (DFT). We obtain the TAEs of 3366 molecules consisting of up to eight non‑hydrogen atoms from the GDB-9 database at the CCSD(T)/CBS level using the high-level W1-F12 theory. This represents the largest and most diverse database of TAEs with confident sub-chemical accuracy for organic molecules. This database is used to assess the performance of 32 DFT functionals. The best-performing method from each rung of Jacob's Ladder is (mean-absolute-deviation in parentheses): pure-GGA B97-D (10.0), meta-GGA B97M-V (2.9), hybrid-GGA CAM-B3LYP-D4 (4.0), and hybrid-meta-GGA M06-2X (1.8 kcal mol⁻¹).