We consider two mechanisms for the production of light clusters: the coalescence model and the thermal model. The first one postulates that light nuclei are formed only at late times of the fireball evolution by recombination of protons and neutrons with close positions and velocities on the kinetic freeze-out surface. On the other hand, the thermal model assumes that deuterons are chemically equilibrated with other hadrons already at hadronisation and
their number does not change during cooling. Then the thermal model describes yields of all hadron species with the universal temperature of T = 156 MeV at LHC. This is very surprising because it is hard to imagine that loosely bound sizeable nuclei can exist in the hot and dense hadron gas. From previous studies, we know that both models predict similar deuteron yields. We try to understand the cluster production and to distinguish between the models also with the help of the anisotropic flow of the clusters.