We present an ab initio potential energy surface for the ground electronic state of ozone. It is global, i.e., it covers the three identical C2v (open) minima, the D3h (ring) minimum, as well as the O(3P)+O2(3Σg−) dissociation threshold. The electronic structure calculations are performed at the multireference configuration interaction level with complete active space self-consistent-field reference functions and correlation consistent polarized quadruple zeta atomic basis functions. Two of the O–O bond distances, R1 and R2, and the O–O–O bending angle are varied on a regular grid (ca. 5000 points with R1⩾R2). An analytical representation is obtained by a three-dimensional cubic spline. The calculated potential energy surface has a tiny dissociation barrier and a shallow van der Waals minimum in the exit channel. The ring minimum is separated from the three open minima by a high potential barrier and therefore presumably does not influence the low-temperature kinetics. The dissociation energy is reproduced up to 90% of the experimental value. All bound states of nonrotating ozone up to more than 99% of the dissociation energy are calculated using the filter diagonalization technique and employing Jacobi coordinates. The three lowest transition energies for O316 are 1101.9 cm−1 (1103.14 cm−1), 698.5 cm−1 (700.93 cm−1), and 1043.9 cm−1 (1042.14 cm−1) for the symmetric stretch, the bending, and the antisymmetric stretch modes, respectively; the numbers in parentheses are the experimental values. The root-mean-square error for all measured transition energies for O316 is only 5 cm−1. The comparison is equally favorable for all other isotopomers, for which experimental frequencies are available. The assignment is made in terms of normal modes, despite the observation that with increasing energy an increasing number of states acquires local-mode character. At energies close to the threshold a large fraction of states is still unambiguously assignable, particularly those of the overtone progressions. This is in accord with the existence of stable classical periodic orbits up to very high energies.