The background of this research found several problems including the use of learning models that were less varied, causing the learning outcomes of some students in mathematics to be low. In addition, the learning media used are less innovative. The purpose of this study was to improve learning outcomes and mathematics learning activities on subtraction material through the Problem Based Learning model assisted by Takberperang media (Subtraction Counting Box) in class IA students at SDN Sampangan 02 Semarang. The type of research used is collaborative Classroom Action Research (PTK) which consists of two cycles, namely cycle I and cycle II. The subjects of this study were class IA students at SD Negeri Sampang 02 Semarang in the 2022/2023 academic year. Data collection techniques used are tests and non-tests (observation). The results showed that in the first cycle the average score was 69, the classical learning mastery was 68%, and the average student learning activity was 80% in the good category. While in cycle II the average score was 81, the classical learning mastery was 82%, and the average student learning activity was 85% in the very good category. This shows an increase from cycle I to cycle II. Based on the results of this study, it was concluded that the Problem Based Learning model assisted by Takberperang media (Calculated Subtraction Box) can improve learning outcomes and mathematics learning activities in class IA students at SDN Sampangan 02 Semarang. The suggestion of this research is that teachers should use a variety of learning models because they can improve learning outcomes, motivation, and student activity in the learning process.