Pairing the distillation stuff with chip export controls discussion is a losing battle. Export controls， and Kimi K3's base model is a way bigger deal -- I personally am skeptical of evidence that Fable was distilled to train K3 directly. Timelines don't match up.
We have information that Moonshot AI distilled Anthropic's Fable for the development of its K3 model. To do this they developed a sophisticated internal platfor...