Low temperature properties of glasses are successfully described by the two level system (TLS) model. As long as the TLS model is valid, TLS relaxations can be classified as equilibrium or nonequilibrium. A question, which is not clearly answered up to now is where the limit of TLS model applicability is and how the deviations from the TLS model manifest themselves in experiments. Measurements of nonequilibrium relaxations via optical hole burning spectroscopy can help to answer this question. It is shown that relaxations in polymer glasses at temperatures above some boundary, which is, for example, about 4K for polymethylmethacrylate and polysrtyrene, requires a new model description.