The Grit scale is a popular measure of achievement-striving behavior. Consisting of two subscales, Consistency of Interests (CI) and Perseverance of Effort (PE), this scale has been repeatedly demonstrated to have high reliability and validity. At the same time, an increasing number of studies explicitly report a low correlation between the subscales and distinct patterns of associations with external measures that each subscale forms. We explored whether there is psychometric evidence that a substantive single grit construct underlies the scale. To answer this question, we investigated the scale structure in a more robust framework than the classical test theory and factor analyses could previously provide. The Russian version of the Grit scale was developed and implemented on a representative sample of high school students (n = 2,269), and different models of item response theory (IRT), both unidimensional and multidimensional, were compared to find the best fitting model. The results confirmed that the subscales reflect related but independent constructs rather than the whole grit construct. The psychometric properties of the subscales were analyzed with the two-dimensional Partial Credit Model. Both subscales of the Russian version of the Grit scale are unidimensional, have good psychometric properties, and can be used to estimate respondents’ ability.