This meta-analysis study aimed to comprehensively examine the effect size of STEM practices on the development of students' problem-solving skills. By combining the results of 25 individual studies, the researchers can gain a holistic understanding of the overall impact of STEM approaches. Consequently, these studies were categorized according to four key characteristics: education levels, teaching methods and techniques, topics, and durations. The analysis revealed that the number of studies with a very high effect size is exactly 11 out of 25, and the average effect size across all studies was also very high. When examining the effect sizes by educational levels, the largest impacts were found for preschool (2.195) and high school (1.673) students, compared to lower effect sizes for middle school (0.645), higher education (0.623), and elementary school (0.506) students. In terms of teaching methods and techniques, the highest average effect sizes were associated with experiential learning (5.2), the design thinking model (3.033), and approaches focused on the earth's layers and disaster-related topics (2.949). In contrast, lower effect sizes were found for the 5E learning model (- 0.346) and content related to electrical conduction (-0.762). The durations of STEM interventions also appeared to influence their effectiveness. Studies with the longest durations of 6-7 weeks (2.960) and 14-16 weeks (1.010) demonstrated greater impacts on problem-solving skills compared to shorter 4-5 week (-0.210) and 8-10 week (0.743) interventions. These findings suggest that STEM practices can have a substantial positive impact on the development of students' problem-solving skills, with certain approaches, educational levels, and durations demonstrating particularly strong effects. The results offer valuable insights for educators and researchers seeking to optimize STEM-based learning experiences.