Jina Reranker
Jina AIJina AI's reranker line, from the v1 turbo and tiny English cross-encoders through v2-base-multilingual to the v3 and v3.5 listwise rerankers. v3 is a 0.6B Qwen3-derived model with a "last but not late interaction" architecture that attends across query and up to 64 documents inside one 131K-token window, reporting 61.94 nDCG@10 on BEIR. The weights are published on the Hub, but under a non-commercial licence.
Openness
2 high confidence- weights
- open(ungated safetensors on the Hub for v1 through v3.5)
- data
- closed(no training data or construction scripts released)
- code
- partial(remote inference code shipped with the card
- license
- Apache-2.0(OSI, and only the superseded v1-turbo-en and v1-tiny-en)+CC-BY-NC-4.0(non-commercial
Jina Reranker's current v2, v3 and v3.5 checkpoints are licensed CC-BY-NC-4.0, which forbids commercial use beyond a narrow AWS and Azure listing without contacting sales; only the older v1-turbo-en and v1-tiny-en checkpoints carry Apache-2.0.
- https://huggingface.co/api/models/jinaai/jina-reranker-v2-base-multilingual recorded 2026-09-11
license cc-by-nc-4.0 recorded in cardData for jinaai/jina-reranker-v2-base-multilingual, confirming the non-commercial terms are not unique to v3.
- https://huggingface.co/jinaai/jina-reranker-v3/raw/main/README.md recorded 2026-09-11
Front matter reads license: cc-by-nc-4.0; the License section says the model is licensed under CC BY-NC 4.0 for use beyond the AWS and Azure listings and directs commercial usage inquiries to contact sales. Weights are ungated, inference is via trust_remote_code with no training pipeline or corpus named.
Adoption
4 high confidenceAdoption is measured as Hugging Face downloads across the text-only reranker line, most of them for v2-base-multilingual and v3. The multimodal m0 reranker is scored as a separate product under the category's vision-language rule, and including its downloads would not change the reading.
- https://huggingface.co/api/models/jinaai/jina-reranker-v2-base-multilingual recorded 2026-09-11
downloads: 1155086; gated: false; private: false for jinaai/jina-reranker-v2-base-multilingual.
- https://huggingface.co/api/models/jinaai/jina-reranker-v3 recorded 2026-09-11
downloads: 780543; gated: false; private: false for jinaai/jina-reranker-v3.
Capability
4 high confidenceJina Reranker v3 scores 61.94 BEIR nDCG@10, level with Qwen3-Reranker-4B's 61.16 on the same table despite using a sixth of the parameters, and both are multilingual with a long context. The comparison comes from Jina's own published table, the only one that reports both models together.
- https://huggingface.co/jinaai/jina-reranker-v3/raw/main/README.md recorded 2026-09-11
Card reports jina-reranker-v3 at 61.94 BEIR, 66.83 MIRACL, 67.92 MKQA and 70.64 CoIR at 0.6B parameters, in a table that also lists Qwen3-Reranker-4B at 61.16 BEIR and bge-reranker-v2-m3 at 56.51; describes 64 documents within a 131K context window.
Verified 2026-09-11