Collisions of6 Li2 molecules with free 6 Li atoms reveal a striking deviation from universal predictions based on long-range van der Waals interactions. Li2 closed-channel molecules are formed in the highest vibrational state near a narrow Feshbach resonance, and decay via two-body collisions with Li2, Li, and Na. For Li2+Li2 and Li2+Na, the decay rates agree with the universal predictions of the quantum Langevin model. In contrast, the rate for Li2+Li is exceptionally small, with an upper bound ten times smaller than the universal prediction. [1,[3][4][5], such as controlling collisions through magnetic/electric field tuning to access scattering resonances or level crossings, understanding whether reactions can depend sharply on the specific quantum state of the collision partners or on the details of short-range inter-particle interactions, characterizing the outgoing states of reaction products, demonstrating coherent control of reaction cross sections, and more.There exists a simple, universal description for twobody inelastic collisions and chemical reactions of an ultracold molecule with another molecule or atom [6][7][8]. This quantum Langevin model assumes a large number of available exit channels in the short-range part of the interaction potential, leading to a unit probability of loss there, and leaving the decay rate dependent on only the long-range van der Waals interaction between collision partners. It has been validated in various experimental settings These universal collisions have a 100% probability of loss at short-range and therefore do not depend on details of the interaction potential there, such as scattering resonances or reactivity determined by matrix elements between quantum states. From a chemistry standpoint, it is thus more interesting to search for examples of collisions that deviate from universality. Such deviations should be more prominent in systems with low mass and consequently a low density of available decay states [14,15], making 6 Li 2 , consisting of the lightest alkali atoms, a uniquely suitable experimental system. We observe that two-body collisions of Li 2 in the highest vibrational state with free Li atoms deviates sharply from universality, as reflected in an exceptionally small two-body decay coefficient. In contrast, the rates for both Li 2 +Li 2 and Li 2 +Na collisions are universal. A recent experiment inferred the rate of Li 2 +Li decay from atomic three-body loss, but in a model-dependent way with uncertainty overlapping both our measurement and the universal prediction [16].To our knowledge this is the first experimental realization of collisions with ultracold molecules where loss is described by physics beyond universal long-range van der Waals interactions [17]. Earlier work by the Rice group reported a low decay rate for Li 2 +Li 2 collisions [18], many orders of magnitude smaller than the universal prediction. In contrast, our measurement demonstrates that this rate is universal, addressing a puzzle that has been prominent for the last decade.We sympa...