Jina: Jina Reranker V3

jina/jina-reranker-v3

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.

Model specifications

Input
text
Output
text
Context
400 tokens
Max output
10,000 tokens
Input price
$0.05 / 1M tokens
Output price
$0.05 / 1M tokens
Released
2026-02-14

Capabilities

  • Streaming
  • Playground

Provider pricing, discounts and data privacy

Compare effective provider prices, published discounts, regions, retention policies, training use, compliance, and privacy links by service tier.

Standard service tier

1 available provider · tier input average $0.05 / 1M tokens · tier output average $0.05 / 1M tokens

Jina

Tier: Standard · Region: US

Pricing
Input
$0.05 / 1M tokens
Output
$0.05 / 1M tokens

No provider discount is currently published.

Data privacy and compliance
Region
US
Zero data retention
No
Data retention
Unknown retention
Used for training
Unknown
Data collection
Moderated
No
GDPR compliant
No
HIPAA compliant
No
SOC 2 certified
No
BYOK supported
No

Frequently asked questions

What is Jina: Jina Reranker V3?
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.
How much does Jina: Jina Reranker V3 cost?
Input costs start at $0.05 / 1M tokens and output costs start at $0.05 / 1M tokens. Provider-level prices vary by service tier.
What is the context length of Jina: Jina Reranker V3?
Jina: Jina Reranker V3 supports a 400 token context window and up to 10,000 output tokens.
What capabilities does Jina: Jina Reranker V3 support?
Jina: Jina Reranker V3 supports Streaming, Playground.
Which providers offer Jina: Jina Reranker V3?
Jina: Jina Reranker V3 is available from Jina.
How do providers handle data privacy for Jina: Jina Reranker V3?
1 of 1 providers report that customer data is not used for training, and 0 offer zero-data-retention routing. Retention, compliance, and privacy-policy links are listed per provider.

Browse all AI models