This research aims to evaluate the quality of an instrument used to measure the chemistry proficiency of senior high school students in Yogyakarta city in response to high order thinking test questions. The evaluation employs the item response theory (IRT) approach, which considers both dichotomous and polytomous data types. The assessment involved analyzing three fundamental assumptions: unidimensionality, local independence, and invariance. The unidimensionality assumption was examined through factor analysis, revealing that the first factor contributed the most to both dichotomous and polytomous data. Local independence was tested using variance-covariance matrix analysis, which revealed that the covariance values between ability intervals were minimal or nearly zero, thereby fulfilling this assumption. Furthermore, invariance testing was conducted on both item parameters and student ability parameters. The results indicated a high correlation (0.893) for item parameters and a moderate correlation (0.761) for ability parameters, confirming that the invariance assumption was satisfied. The students’ ability levels were categorized into five levels, ranging from very low to very high. The analysis revealed that dichotomous data covered a broader range of abilities than polytomous data did. Overall, the study confirms that the instrument used is valid, reliable, and aligns with the core assumptions of IRT. These findings have significant implications for enhancing the quality of chemistry assessment tools using the IRT approach, ensuring that test questions are more accurate and dependable. These insights can help teachers design instructional methods that promote higher-order thinking skills (HOTSs), encourage the broader application of IRT in assessment, and support data-driven educational policies. Moreover, students gain improved training in analytical thinking and problem solving, leading to a deeper understanding of chemistry.