The purpose of the study was to improve mathematical problem solving for middle school students with learning disabilities by implementing a research-based instructional program in inclusive general education math classes. A total of 40 middle schools in a large urban district were matched on state assessment performance level (low, medium, and high performing) and socioeconomic status. One school from each pair was randomly assigned to the intervention condition, and one eighth grade math teacher participated at each school (n = 40). Because of attrition at the outset, 24 schools completed the study (8 intervention, 16 comparison). The intervention, "Solve It!", a research-based cognitive strategy instructional program, was implemented for 7 months, and periodic progress monitoring was conducted. A cluster-randomized design was used, and the data were consistent with a three-level model in which repeated measures were nested within students and students were nested within schools. The results indicated that students who received the intervention (n = 319) showed significantly greater growth in math problem solving over the school year than students in the comparison group (n = 460) who received typical classroom instruction. Moreover, the intervention effects did not differ for students with learning disabilities, low-achieving students, and average-achieving students. Thus, the findings were positive and support the efficacy of the intervention when implemented by general education math teachers in inclusive classrooms. (Contains 3 tables and 4 figures.)