This paper presents the results of a quantitative study investigating the development of teachers' attitudes and self-efficacy expectations for inclusive mathematics instruction in the context of an in-service training that was designed in a blended learning format and compared to an unsupported online offer. In the blended learning format, 101 teachers participated in an in-service training, designed as a combination of six workshops with concrete activities based on materials for inclusive mathematics teaching and intermittent practical phases with collaborative learning environments for children aged 6-12 years. The teachers' attitudes toward inclusive mathematics instruction and self-efficacy expectations are collected by using published scales (KIESEL) and scales under development. The effects of the blended learning program are analyzed by the t test for dependent samples or the nonparametric Wilcoxon signed-rank test and compared to the effects of an online offer without further guidance and support. On the one hand, the results indicate the importance of the blended learning program in comparison with the unsupported online offer. On the other hand, they show insights into the relevance of the participants' interaction with the materials during the concrete activities as well as of their try outs in and their reflection on teaching practice during the intermittent practical phases.