Students' proficiency in the language of instruction is essential for their mathematical learning. Accordingly, language-responsive instruction, which includes adapting teaching material to students' language needs, is thought to promote mathematical learning, particularly for students with lower levels of proficiency in the language of instruction. However, empirical evidence for the effectiveness of this type of instruction in heterogeneous classrooms is scarce, and potential differential effects for learners with different learning prerequisites still need to be studied. The present study examines whether language-responsive instructional materials can promote students' learning of fractions. We conducted a quasi-experimental intervention study with a pre- and posttest in Grade 7 (N = 211). The students were assigned to one of three instructional conditions: fraction instruction with or without additional language support or to a control group. The results showed that both intervention groups had higher learning gains than the control group. However, students with lower proficiency in the language of instruction benefited more from fraction instruction with additional language support than without it. The opposite was true for students with higher proficiency in the language of instruction. Moreover, learning gains depended on students' levels of mathematics anxiety. Our study contributes to a more detailed understanding of the effectiveness of language-responsive instruction in heterogeneous classrooms.