Based on public literature, as of October 2026 it is difficult to conclude that distillation is a major factor in the performance of reasoning models