This study aims to analyze the improvement of students' mathematical problem-solving abilities through the implementation of the AIR learning model with the RME approach on the topic of flat sided geometric material in Grade VIII at SMP Negeri 45 Batam. This research is a quasi-experimental type of research, with a pretest-posttest control group design. The sample was selected using probability sampling with a cluster random sampling method, so that two sample classes were obtained, VIII C as the experimental class and VIII F as the control class. The test instrument in the form of essay questions to measure students' mathematical problem solving abilities on flat-sided geometric material was used to collect research data. The research results showed that the average improvement in mathematical problem-solving abilities for the experimental class was 0.42, while for the control class it was 0.31. From the Independent t-test of N-Gain, the Sig. value was 0.044, it can be concluded that the average mathematical problem solving ability of students through the application of the AIR learning model with the RME approach is higher than the average mathematical problem solving ability of students using conventional learning models.