Mathematical modeling is one of the essential 21st-century skills that students must possess to connect mathematical concepts with real-world problems. Despite its recognized importance, the performance of Indonesian students in modeling tasks remains low, as reflected in the 2022 PISA results. This study aims to analyze students' mathematical modeling ability in solving scientific-context problems derived from the 2022 PISA-released items. The research used a descriptive qualitative method involving 25 seventh-grade students from SMP IT Raudhatul Ulum. Data were collected through students' written responses to four PISA problems and analyzed using a rubric based on modeling indicators by Blum and Leiss. The findings show that most students—56%—are in the low category of modeling ability, 36% are in the medium category, and only 8% reach the high category. Students struggle to translate real-world problems into mathematical models and interpret results contextually. These findings highlight the need for contextual learning strategies to strengthen students' modeling competencies in scientific settings.