Existing approaches for training paraeducators rely heavily on intensive one-to-one coaching and may not be feasible in practice. In this study, we test a tiered training model in which all paraeducators first received group training, and then coaching was provided only for the subset who did not meet performance criteria after group training. Using a concurrent multiple-probe design staggered across classrooms, we demonstrated a functional relation between the tiered model and implementation fidelity of two systematic prompting strategies across 13 paraeducators in five schools. Nine paraeducators achieved the performance criterion for both practices with group training alone, and the remaining four met the criterion after teacher-delivered coaching. In addition, paraeducators generalized implementation to new situations, and students with severe disabilities who received instruction made progress on individualized goals. Based on these findings, a tiered training model is a feasible and promising means to train paraeducators.