jina-reranker-v3
0.6B multilingual listwise reranker for RAG pipelines.
About
jina-reranker-v3 is a 0.6B-parameter multilingual document reranker from Jina AI built on Qwen3-0.6B, using a last-but-not-late interaction design that reranks up to 64 documents in a single 131K-token inference call. It reaches 61.94 nDCG@10 on BEIR — state of the art at its size and 10x smaller than generative listwise rerankers.
It is open-weight under CC-BY-NC-4.0, so commercial use is not covered, and GGUF/MLX builds are available for local deployment.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
BEIR 61.94 nDCG@10; MIRACL 66.83 (Jina-reported).
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 7/10
Scored 7/10 (verified tier, metrics-v1). Evidence: SOTA BEIR at 0.6B; cross-document 64-doc single-call reranking; CC-BY-NC-4.0 limits commercial use; GGUF/MLX builds available; paper arXiv 2509.25085. License: CC-BY-NC-4.0 per the Hugging Face model card metadata..
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 7.0/10
- Maintenance · recency, releases, issue load
- 7.0/10
- Openness · license + docs
- 6.0/10
- Manual review · sanity read + red-flag check
- 7.0/10
- · SOTA BEIR at 0.6B; cross-document 64-doc single-call reranking
- · CC-BY-NC-4.0 limits commercial use; GGUF/MLX builds available
- · paper arXiv 2509.25085