Reranker

Maximize the search relevancy and RAG accuracy with our cutting-edge reranker API.

Reranker API

Try our cutting-edge reranker API to maximize your search relevancy and RAG accuracy. Starting for free!
Number of returned documents
top_n
top_k
Maximum number of top-ranked documents to return.
Request
POST
curl "https://api.jina.ai/v1/rerank" \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $JINA_API_KEY" \
  -d @- <<EOFEOF
{
  "model": null,
  "query": "Organic skincare products for sensitive skin",
  "top_n": 3,
  "documents": [
    "Organic skincare for sensitive skin with aloe vera and chamomile: Imagine the soothing embrace of na…ing, healthy complexion.",
    "New makeup trends focus on bold colors and innovative techniques: Step into the world of cutting-edg…atement with every look.",
    "Bio-Hautpflege für empfindliche Haut mit Aloe Vera und Kamille: Erleben Sie die wohltuende Wirkung u…einen strahlenden Teint.",
    "Neue Make-up-Trends setzen auf kräftige Farben und innovative Techniken: Tauchen Sie ein in die Welt…jedes Mal ein Statement.",
    "Cuidado de la piel orgánico para piel sensible con aloe vera y manzanilla: Descubre el poder de la n…el radiante y saludable.",
    "Las nuevas tendencias de maquillaje se centran en colores vivos y técnicas innovadoras: Entra en el … y destaca en cada look.",
    "针对敏感肌专门设计的天然有机护肤产品：体验由芦荟和洋甘菊提取物带来的自然呵护。我们的护肤产品特别为敏感肌设计，温和滋润，保护您的肌肤不受刺激。让您的肌肤告别不适，迎来健康光彩。",
    "新的化妆趋势注重鲜艳的颜色和创新的技巧：进入化妆艺术的新纪元，本季的化妆趋势以大胆的颜色和创新的技巧为主。无论是霓虹眼线还是全息高光，每一款妆容都能让您脱颖而出，展现独特魅力。",
    "敏感肌のために特別に設計された天然有機スキンケア製品: アロエベラとカモミールのやさしい力で、自然の抱擁を感じてください。敏感肌用に特別に設計された私たちのスキンケア製品は、肌に優しく栄養を与え、保護します。肌トラブルにさようなら、輝く健康な肌にこんにちは。",
    "新しいメイクのトレンドは鮮やかな色と革新的な技術に焦点を当てています: 今シーズンのメイクアップトレンドは、大胆な色彩と革新的な技術に注目しています。ネオンアイライナーからホログラフィックハイライターまで、クリエイティビティを解き放ち、毎回ユニークなルックを演出しましょう。"
    + add input
  ],
  "return_documents": false
}
EOFEOF

v3.5: Faster Listwise Reranking with Hybrid Attention and Self-Distillation

jina-reranker-v3.5 is a 0.6B parameter multilingual listwise reranker with a 131K context window. Hybrid attention and self-distillation make it both faster and more robust across domains than v3, and it is a drop-in replacement: the request schema is unchanged.

m0: Multilingual Multimodal Document Reranker

Our new multimodal multilingual reranker for retrieving visual documents across multiple languages, with SOTA performance on multilingual long documents and code searching tasks.

What is a Reranker?

The goal of a search system is to find the most relevant results quickly and efficiently. Traditionally, methods like BM25 or tf-idf have been used to rank search results based on keyword matching. Recent methods, such as embedding-based cosine similarity, have been implemented in many vector databases. These methods are straightforward but can sometimes miss the subtleties of language, and most importantly, the interaction between documents and a query's intent. This is where the "reranker" shines. A reranker is an advanced AI model that takes the initial set of results from a search—often provided by an embeddings/token-based search—and reevaluates them to ensure they align more closely with the user's intent. It looks beyond the surface-level matching of terms to consider the deeper interaction between the search query and the content of the documents.
Here's how it works:
1
Initial Retrieval
A search system uses embeddings/BM25 to find a broad set of potentially relevant documents based on the user's query.
2
Reranking
The reranker then takes these results and analyzes them at a more granular level, considering the nuances of how the query terms interact with the document content.
3
Improved Results
It reorders the search results, placing the ones it deems most relevant at the top, based on this deeper analysis.
The reranker can significantly improve the search quality because it operates at a sub-document and sub-query level, meaning it looks at the individual words and phrases, their meanings, and how they relate to each other within the query and the documents. This results in a more precise and contextually relevant set of search results.
Jina Reranker v2, released in June 2024, was built for Agentic RAG: function-calling support, multilingual retrieval across more than 100 languages, and code search. It remains available, though jina-reranker-v3.5 supersedes it for text reranking. Read more about the v2 model.
Multilingual Retrieval
Reranker v2 enables document retrieval in over 100 languages, regardless of the query language.
Function-Calling & Code Search
Reranker v2 ranks code snippets and function signatures based on natural language queries, ideal for Agentic RAG applications.
Tabular and Structured Data Support
Reranker v2 ranks the most relevant tables based on natural language queries, helping to sort different table schemas and identify the most relevant one before generating an SQL query.

Two Ways to Purchase

Subscribe to our API or purchase through cloud providers.
With Jina Search Foundation API
The easiest way to access all of our products. Top-up tokens as you go.
Top up this API key with more tokens
Depending on your location, you may be charged in USD, EUR, or other currencies. Taxes may apply.
Please enter the correct API key to top up.

On-premises deployment

Deploy Jina Reranker on AWS SageMaker and Microsoft Azure and soon in Google Cloud Services, or contact our sales team to get customized Kubernetes deployments for your Virtual Private Cloud and on-premises servers.
AWS SageMaker
Embeddings
Reranker
Microsoft Azure
Embeddings
Reranker
Google Cloud
Embeddings

Comparison of Reranker, Vector Search, and BM25

The table below provides a comprehensive comparison of the Reranker, Vector/Embeddings Search, and BM25, highlighting their strengths and weaknesses across various categories.
RerankerVector SearchBM25
Best ForEnhanced search precision and relevanceInitial, rapid filteringGeneral text retrieval across wide-ranging queries
GranularityDetailed: Sub-document and query segmentBroad: Entire documentsIntermediate: Various text segments
Query Time ComplexityHighMediumLow
Indexing Time ComplexityNot requiredHighLow, utilizes pre-built index
Training Time ComplexityHighHighNot required
Search QualitySuperior for nuanced queriesBalanced between efficiency and accuracyConsistent and reliable for a broad set of queries
StrengthsHighly accurate with deep contextual understandingQuick and efficient, with moderate accuracyHighly scalable, with established efficacy
Try reranker API for freeTry embedding API for free

Learning about Reranker

What is a reranker? Why is vector search or cosine similarity not enough? Learn about rerankers from the ground up with our comprehensive guide.
August 03, 2026 • 11 minutes read
jina-reranker-v3.5: Faster Listwise Reranking with Hybrid Attention and Self-Distillation
A 0.6B listwise reranker that beats Qwen3-Reranker-4B on BEIR, reranks up to 1.56x faster than v3, and gains 9.6 nDCG@10 on semi-structured retrieval.
July 20, 2026
jina-reranker-v3.5: An Efficient Listwise Reranker with Hybrid Attention and Self-Distillation
Listwise rerankers are the discriminative core of agentic retrieval pipelines, yet production deployment demands efficiency, domain robustness, and fluency on semi-structured data at the same time. We present jina-reranker-v3.5, a 0.6B-parameter listwise reranker that meets these demands together without sacrificing the cross-document comparison that makes its predecessor jina-reranker-v3 effective. It replaces uniform global attention with a hybrid schedule of three sliding-window layers followed by two global layers, pinning the terminal layer to global as last-but-not-late readout requires. It trains on a curated multi-domain mixture spanning legal, medical, financial, multilingual, and structured retrieval, and transfers quality through a three-stage self-distillation recipe in which a full-attention teacher sets an upper bound that a sparse-attention student then recovers. jina-reranker-v3.5 reaches 63.20 nDCG@10 on BEIR, matching a 4B model at roughly 7x fewer parameters, and lifts semi-structured retrieval by 9.6 nDCG@10 over jina-reranker-v3 while cutting listwise inference latency by up to 1.56x.
October 03, 2025 • 7 minutes read
Jina Reranker v3: 0.6B Listwise Reranker for SOTA Multilingual Retrieval
New 0.6B-parameter listwise reranker that considers the query and all candidate documents in a single context window.
October 01, 2025
jina-reranker-v3: Last but Not Late Interaction for Document Reranking
jina-reranker-v3 is a 0.6B parameter multilingual document reranker that introduces a novel last but not late interaction. Unlike late interaction models such as ColBERT that perform separate encoding followed by multi-vector matching, our approach conducts 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. This compact architecture achieves state-of-the-art BEIR performance with 61.94 nDCG@10 while being significant smaller than generative listwise rerankers.

FAQ

How to get my API key?

What's the rate limit?

Rate limit
Rate limits are tracked in two ways: RPM (requests per minute) and TPM (tokens per minute). Limits are enforced per IP/API key and will be triggered when either the RPM or TPM threshold is reached first. When you provide an API key in the request header, we track rate limits by key rather than IP address.
ProductAPI EndpointDescriptionw/o API Keyw/ Free API Keyw/ Paid API Keyw/ Premium API KeyAverage latencyToken Usage CountingAllowed Request
Reader APIhttps://r.jina.aiConverts a URL to LLM-friendly text20 RPM500 RPM500 RPM5000 RPM7.9sCount the number of tokens in the output response.GET/POST
Reader APIhttps://s.jina.aiSearch the web and convert results to LLM-friendly text100 RPM100 RPM1000 RPM2.5sEvery request costs a fixed number of tokens, starting from 10,000 tokensGET/POST
Embedding APIhttps://api.jina.ai/v1/embeddingsConvert text/images to fixed-length vectors100 RPM & 100,000 TPM500 RPM & 2,000,000 TPM5,000 RPM & 50,000,000 TPM
depends on the input size
Count the number of tokens in the input request.POST
Reranker APIhttps://api.jina.ai/v1/rerankRank documents by query100 RPM & 100,000 TPM500 RPM & 2,000,000 TPM5,000 RPM & 50,000,000 TPM
depends on the input size
Count the number of tokens in the input request.POST

Do I need a commercial license?

CC BY-NC License Self-Check
Are you using our hosted API, or our official images on Azure, AWS, or GCP?
API-related common questions
Billing-related common questions
Current language / theme
Search Foundation
Reader
Embeddings
Reranker
Get Jina API key
Rate limit
About us
News
Download Jina logo
Download Elastic logo
API Status
Elastic © 2026.SecurityTerms & ConditionsPrivacyManage CookiesDo Not Sell or Share My Personal Information
This website and all associated content, software, products, and services are intended for professional use only. No consumer use is intended or directed.