Learning environments deserve optimal acoustics to support speech communication, especially at the first stages of schooling. To guarantee an adequate sound environment for educational facilities, the control of background noise and the reduction late reverberation constitute two of the main intervention strategies at the designing stage. Both in the case of new construction and renovation of schools, prediction models for speech intelligibility can be used as evaluation tools to assess the efficacy of acoustic treatments for the enhancement of the speech intelligibility perceived by listeners. This work presents the results of the application of two versions of the Binaural Speech Intelligibility Model (BSIM) to a real primary school classroom that underwent an acoustical renovation. The BSIM versions mainly differed in the pre-processing of the speech signal, where the energy of the late reflections was either considered as detrimental to speech intelligibility or not. The acoustic renovation complied with the UNI 11532-2 standard, as objective measures of reverberation time decreased while speech clarity and definition increased, as expected. The BSIM gave outcomes in terms of speech recognition thresholds, considering the effect of increasing talker-to-listener distance and of the binaural spatial release from masking due to the separation of noise- and speech-sources.