Model reference · open weights

sup-simcse-ja-large

sup-simcse-ja-large is an open-weight embedding model from cl-nagoya. sup-simcse-ja-large (BF16) weighs 1.3 GB; the smallest configuration that runs it is RTX 3060 12 GB.

• sup-simcse-ja-large is a Japanese feature-extraction model developed by cl-nagoya for generating sentence embeddings.
• It is based on the cl-tohoku/bert-large-japanese-v2 architecture and was fine-tuned using the Supervised SimCSE method on the JSNLI dataset.
• The model supports a context length of 512 tokens and is distributed under the cc-by-sa-4.0 license.

Summary of the cl-nagoya/sup-simcse-ja-large model card, 2026-10-01

Hugging Face ↗