The present paper analyzes the impact of blended learning (BL) on the academic achievement of higher education students. A meta-analysis (k = 51 effect sizes) was conducted to perform a statistical synthesis of studies contrasting student performance in BL conditions with traditional classroom instruction. We include disciplines and instructors’ end-of-course evaluation method as moderating variables. The results show that BL demonstrates a small summary effect (g+ = 0.385, p < 0.001) compared to traditional teaching methods A significantly higher mean effect size was found in STEM disciplines (g+ = 0.496) compared to that of non-STEM disciplines (g+ = 0.210). Nevertheless, the weighted mean effect sizes reveal no significant differences regarding of end-of-course assessment methods, namely one-moment and multiple-component assessment. The finding confirms that BL is significantly associated with greater learning performance of STEM-disciplined students than with traditional classroom practice. Accordingly, discussion concerning the findings and implications for future research are elaborated.