jina-reranker-v3 is a 0.6B parameter multilingual document reranker introducing a novel last but not late interaction architecture. Unlike ColBERT's separate encoding with multi-vector matching, this model performs causal self-attention between query and documents within the same context window, enabling rich cross-document interactions before extracting contextual embeddings from the last token of each document. Built on Qwen3-0.6B with 28 transformer layers and a lightweight MLP projector (1024→512→256), it processes up to 64 documents simultaneously within 131K token context. The model achieves state-of-the-art BEIR performance with 61.94 nDCG-10 while being 10× smaller than generative listwise rerankers.
https://api.jina.ai/rerank| Model | Context | Max out | Pricing |
|---|---|---|---|
| Jina Reranker v1 Base EN | 8K | — | $0.05— |
| Jina Reranker v1 Tiny EN | 8K | — | $0.05— |
| Jina Reranker v1 Turbo EN | 8K | — | $0.05— |
| Jina Reranker v2 Base Multilingual | 1K | — | $0.05— |
| Jina Reranker v3.5 | 131K | — | $0.05— |
| Jina Reranker v3 | 134K | — | $0.05— |
/v1/models/jina/jina-reranker-v3jina-reranker-v3 is a 0.6B parameter multilingual document reranker introducing a novel last but not late interaction architecture. Unlike ColBERT's separate encoding with multi-vector matching, this model performs causal self-attention between query and documents within the same context window, enabling rich cross-document interactions before extracting contextual embeddings from the last token of each document. Built on Qwen3-0.6B with 28 transformer layers and a lightweight MLP projector (1024→512→256), it processes up to 64 documents simultaneously within 131K token context. The model achieves state-of-the-art BEIR performance with 61.94 nDCG-10 while being 10× smaller than generative listwise rerankers.