Model reference · open weights

splade

splade is an open-weight embedding model from naver. splade-v3 (BF16) weighs 438 MB; the smallest configuration that runs it is RTX 3060 12 GB.

• SPLADE-v3 is a feature-extraction model developed by Naver for semantic search and sparse retrieval.
• It maps English text to a 30522-dimensional sparse vector space using a maximum sequence length of 512 tokens.
• The model is distributed under the cc-by-nc-sa-4.0 license.

Summary of the naver/splade-v3 model card, 2026-10-01

Hugging Face ↗Paper ↗