Co-teaching and related collaborative models of instruction have become widely used models in many school systems of high-income countries. However, the overall picture of the effects of collaborative models of instruction on students’ academic achievement has not yet been firmly drawn. This systematic review and meta-analysis aim to collect this picture by investigating the mean effect size and how effects vary across various methodological and theoretical moderators. Although previous research and reviews contend that the evidence base is scarce, we find 128 treatment and control group designed studies in the period from 1983-2020 stored in common social science databases and registers. We excluded 52 studies due to critical risk of bias via Cochrane's risk of bias assessment tools and conducted a meta-analysis of 76 studies yielding 280 short-term effect sizes of which 82 percent are pretest-adjusted effect sizes. We find a moderate statistical significant overall average effect size equal to g ̅=0.11 (95 % "CI" [0.035-0.184]) using the correlated-hierarchical effects model. From subgroup analyses, we find that collaborative instruction generally yields stable moderate effects on academic achievement suggesting an increased potential for the scalability of these interventions. We also find that important efficiency factors highlighted in the co-teaching literature do not explain true variation in effect sizes. Finally, we do not find any clear evidence for publication bias or small study effects. This review is limited by a large number of included quasi-experimental studies. Future research should concentrate on conducting more rigorous experimental research, especially in the field of co-teaching.