This paper aims to explore the construct validity of grammar gap-filling task in National Matriculation English Test using the basic Rasch modeling approach. Fit statistics, measurement invariance, unidimensionality and DIF index were employed to reflect the construct validity. 401 senior three students completed the grammar gap-filling task of 2018 National Matriculation English and the scores were analyzed by Winsteps3.72.3. The research results show that the construct validity of grammar gap-filling task in National Matriculation English can be improved though the fit statistics, measurement invariance, unidimensionality and DIF index support it in general.