Abstract
<title>Abstract</title> <p> Sustainable Development Goal 4 demands inclusive, equitable quality education, yet science, technology, engineering, and mathematics (STEM) inequities persist. In Ethiopia, over 88% of youth lack basic science proficiency, and secondary completion is 28%. This PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) compliant meta-analysis synthesizes 42 randomized controlled trials and quasi-experiments (2018–2026; N = 14,728) comparing blended STEM to traditional instruction using a random-effects model (Hedges' <italic>g</italic> ). Grounded in the Technological Pedagogical Content Knowledge (TPACK) framework (extended with Sustainable Development Goal 4 and social and emotional learning), blended STEM significantly outperformed traditional instruction (g = 0.74, 95% CI [0.59, 0.89], p< .001; I²=76%). Mastery-based flexible pacing models yielded larger effects (g = 1.28; 18 studies). African contexts showed robust effects (g = 0.68; 8 studies), including a 22% literacy gain in Ethiopia (normalized gain = 0.75). Gender equity analysis favored girls (g = 0.79 vs. 0.62 for boys), reducing SDG target 4.5 gaps by 21%. Heterogeneity was partially explained by dosage (r = 0.45, p < .01). No publication bias was detected (Egger's p = .11). GRADE indicated moderate certainty. Blended STEM, especially mastery-based models, effectively advances Sustainable Development Goal targets 4.1, 4.4, 4.5, 4.7, and 4.c. Policy should mandate TPACK certification, integrate social and emotional learning in conflict zones, and scale artificial intelligence offline tools via public-private partnerships. </p>