Modelling instruction is systematic instructional activity for constructing and applying scientific knowledge in Physics lesson. The purpose of this research is to determine the effect of Modelling instruction with different reasoning abilities on understanding physical concepts by controlling students’ prior knowledge. This research used experimental method with 2x2 factorial design with two Modelling instruction classes and two conventional classes with a total of 176 students. The instrument used was reasoning ability test, prior knowledge test, and physics concept test. It used LCTSR (Lawson’s Classroom Test of Scientific Reasoning) instrument. Prior knowledge test instruments consisted of 25 problems to identify how deep the students understand the topic before they undergo the learning process and physics concept test consisted of 25 problems. Based on the statistical test using two factor Ancova, it proved that there was a significant difference in students’ ability to master the physics concept between using Modelling instruction learning model and using conventional learning model. The result showed that the Modelling instruction increasing conceptual understanding better than conventional learning. There are two important parts in the Modelling instruction that are model development and model deployment. This study also confirms that there are significant differences in understanding the concepts between students of high reasoning ability and low reasoning ability. Students with high reasoning abilities have a better understanding of concepts than students with low reasoning abilities.