A reranker for legal and administrative document search, built on Qwen3-0.6B.
- Parameters
- 609M
- License
- cc-by-nc-4.0
- Max sequence length
- 12,288 tokens
- Updated
- Sep 29, 2026
Search foundation models
Browse the catalogue. Select a model for its full specifications and model card.
Reranker
Reorders retrieved passages by relevance to the query, so the strongest evidence reaches the LLM first.
Embeddings
Turns queries and documents into dense vectors for the first stage of retrieval — vector search over your corpus.
In development
Text2Cypher
Translates natural-language questions into Cypher, so a graph database can answer them — the query layer of GraphRAG.
In development