During an internal meeting of ByteDance’s Seed team, Zhang Yiming made it clear that the company would not resort to distillation technology as a shortcut to bolster the capabilities of its AI models. Instead, ByteDance is committed to prioritizing long-term objectives, even if it means sacrificing short-term benefits. As one of China’s leading AI companies, ByteDance predominantly relies on proprietary models. Its steadfast refusal to employ distillation techniques has been viewed by outsiders as a contributing factor to the comparatively slower progress of its large language models. The debate within the team regarding the adoption of distillation technology has persisted for quite some time. This disagreement has only deepened with the swift advancement of open-source models by domestic competitors. Distillation technology offers the advantage of reducing training time and computational resource consumption, enabling new models to directly assimilate the reasoning processes of state-of-the-art models.