Students with and without disabilities may be educated across various service delivery models (SDMs): general education, cotaught, pull-out, and self-contained. Still, evidence for their relative effectiveness at scale remains limited. Using longitudinal administrative data from Indiana, we measured the effect of different SDMs on test scores, attendance, and disciplinary incidents. We leveraged within-student variation in SDM assignments and differences across students, applying student fixed effects and lagged outcomes models, which bound the causal effect within a narrow, policy-relevant range. Students with disabilities performed better in less restrictive environments, although the magnitude was often modest and varied across SDMs. Coteaching had a positive impact on students without disabilities. This work contributes to our understanding of inclusive practices' effectiveness as experienced statewide.