This study aims to describe the implementation of the Team Games Tournament (TGT) learning model based on a reward system and to determine the improvement of learning activity among third-grade students at MI Miftahul Huda Tegalsambi in Mathematics. This research was motivated by the low level of student learning activity due to the conventional or teacher-centered learning methods used by teachers, causing students to be passive, rarely ask questions, and show less enthusiasm during discussions. This study employed Classroom Action Research (CAR) using the Kemmis and McTaggart model, which was conducted in two cycles. Each cycle consisted of planning, action implementation, observation, and reflection stages. The subjects of this study were 20 third-grade students at MI, consisting of 12 male students and 8 female students. Data were collected through observation, interviews, and documentation, then analyzed using descriptive quantitative and qualitative methods. The results showed that the implementation of the TGT learning model based on a reward system significantly improved students’ learning activity. The average score (Mean) of learning activity increased from 78.67 in Cycle I to 94.75 in Cycle II, the Median score increased from 77.50 to 95.00, and the Mode score increased from 75.00 to 96.66. This improvement was supported by several enhancements in Cycle II, including more intensive motivation, variations in rewards, and adjustments to the difficulty level of tournament questions. Therefore, it can be concluded that the TGT learning model based on a reward system is effective in improving students’ learning activity in lower-grade classes of Islamic elementary schools.