This study aims: 1) to analyze the effect of the scientific approach model on improving primary school students' learning outcomes in high and low grades. 2) to analyze the effect of the scientific approach model on improving the learning outcomes of elementary school students based on the model used. This study used a meta-analysis research method with 18 articles as samples from national journals in the 2014-2022 range and the instrument was a sheet with coding. The stages for conducting meta-analysis research are problem formulation; sampling data collection; data analysis, interpretation and presentation of results. The meta-analysis research shows that 1) on the subject of low and high grades, the PjBL (Project Based Learning) model with a scientific approach has a high effect on student learning outcomes in low grades with an average effect size of 0.602 while the inquiry model with a scientific approach has a high effect on student learning outcomes in high grades with an average effect size of 0.525; 2) seen from the type of model used, the PBL (Problem Based Learning) model with a scientific approach has a high effect on improving the learning outcomes of elementary school students with an average effect size of 0.488. It can be concluded in this study that the PjBL (Project Based Learning) model based on the application of a scientific approach is more effective in improving the learning outcomes of elementary school students in high grades, while in low grades the incuri model based on the application of a scientific approach is more effective in improving the learning outcomes of elementary school students in the scientific approach, and the PBL (Problem Based Learning) model based on the application of a scientific approach is more effective in improving the learning outcomes of elementary school students in the model used.