KURE v1 is a specialist — it speaks Korean fluently at the embedding level, producing dense vector representations tuned for Korean-language semantic tasks. It handles nuance in Korean text that general multilingual models often flatten or miss. The trade-off is narrow scope: it outputs embeddings only, and works best when your data is primarily Korean.