Extracting useful knowledge from user reviews of applications, published on commercialization platforms, social networks, and other digital environments, is a new trend in the software support or maintenance process. These reviews constitute a very valuable source of needs, demands, dissatisfaction and ratings about software applications they use. The automatic detection of informative reviews is one of the problems currently being addressed in order to mining this large volume of information. The purpose of this work is to evaluate the performance of Pre-trained Transformers Models for deal with informative opinions filtering problem, in order to identify which of these models provides the best quality results. As part of this work, five Transformers models were studied and evaluated with four test collections, about reviews of four different applications. The results obtained were very promising where BERTweet model achieved the highest efficiency, overcoming the results of other state-of-the-art solutions.