Last released Nov 20, 2026
Distill your LLM classifier into a small fast model, with a statistical bound on agreement and LLM fallback for uncertain cases.