<![CDATA[This research is motivated by the low mathematical problem-solving ability of elementary school students, especially when facing non-routine problems that require complex reasoning. The integration of Realistic Mathematics Education (RME) and Computational Thinking (CT) is seen as an approach capable of presenting contextual learning while stimulating systematic thinking processes. This research uses a quantitative method with a Quasi-Experimental Non-Equivalent Control Group Design involving fifth-grade students from two elementary schools in Prambon District. The experimental group received RME learning integrated with CT, while the control group received direct learning. Data were collected through pre-tests, post-tests, and self-efficacy questionnaires, then analyzed using a two-way ANOVA to examine the effect of the learning model, self-efficacy, and the interaction between the two. The results showed that the RME-CT model significantly improved mathematical problem-solving ability compared to conventional learning, with higher score increases in the experimental group. In addition, self-efficacy was shown to moderate learning effectiveness, where students with high self-efficacy achieved better results. There was a significant interaction between learning models and self-efficacy on learning outcomes. This research contributes to offering an innovative approach that combines realistic contexts and computational thinking as an effective strategy to improve elementary school students' mathematical competence and self-confidence.]]>
