Cognitive flexibility (CF) is the ability to adapt cognitive strategies according to the changing environment. The deficit in CF has often been linked to various neurological and psychiatric disorders including schizophrenia. However, the operationalization and assessment of CF have not been unified and the current research suggests that the available instruments measure different aspects of CF. The main objective of the present study was to compare three frequently used neuropsychological measures of CF-Wisconsin Card Sorting Test (WCST), Trail Making Test (TMT) and Stroop Color and Word Test (SCWT) in a population of patients (N = 220) with first-episode schizophrenia spectrum disorders in order to evaluate their convergent validity. The hypothesis of an underlying latent construct was tested via a confirmatory factor analysis. We used a one-factor CF model with scores from WCST, SCWT and TMT as observed variables. The established model showed a good fit to the data (χ2 = 1.67, p = 0.43, SRMR = 0.02, RMSEA = 0.0, CFI = 1.00). The highest factor loading was found in WCST as CF explained most of the variance in this neuropsychological measure compared to the other instruments. On the other hand, a TMT ratio index and a SCWT interference demonstrated lowest loadings in the model. The findings suggest that not all the frequently used measures share an underlying factor of CF or may capture different aspects of this construct.