Skip to content

Jina AI · Embedding

Jina Embeddings v4

Jina AI 🇩🇪EmbeddingJina Embeddings family

Jina Embeddings V4 is a 3.8 billion parameter multimodal embedding model that provides unified text and image representation capabilities. Built on the Qwen2.5-VL-3B-Instruct backbone, the model features an architecture that supports both single-vector and multi-vector embeddings in the late interaction style, addressing limitations found in traditional CLIP-style dual-encoder models. The model incorporates three specialized task-specific LoRA adapters (60M parameters each) that optimize performance across different retrieval scenarios including asymmetric query-document retrieval, semantic text similarity, and code search without modifying the frozen backbone weights. The model demonstrates strong performance in processing visually rich content such as tables, charts, diagrams, screenshots, and mixed-media formats through a unified processing pathway that reduces the modality gap present in conventional architectures. Supporting multilingual capabilities, the model can handle input texts up to 32,768 tokens with images resized to 20 megapixels, making it suitable for various document retrieval and cross-modal search applications across different languages and domains.

llminfo IQ
114.1
Maker's price
$0.05 /1M
Max context
33K
Listings
1

Evidence and sources

🤗 jinaai/jina-embeddings-v4llminfo IQ 114.1, ranked among embedding models. From 26 benchmarks. Plenty of data behind it.How llminfo IQ is computedReleased 23 Jun 2025Maker HQ: DE (source)
Sources: Jina AI data docs, llminfo researchPrimary data from Jina AI from llminfo research: release date

Advantages and trade-offs

Compared with 10 embedding models, picked mainly by popularity and IQ
We only make a claim when there's enough data on models of the same type. Updated 22 Nov, 04:34 UTC.Prices come from the model's maker where they sell it, otherwise their official partner, otherwise the market median. This comparison uses 5 maker's prices; 6 market median prices.

Input & output

as stated by the model's providers
Text
Images
Embeddings

Listings

 Jina AIRegions: 🇺🇸 USjina-ai/jina-embeddings-v4Source: Jina AI data docs33K$0.05—

Benchmarks

Named effort levels run from lowest to highest. Select a result to see its source details.

Embedding

MLEB12 related benchmarks · 12 results
SCALR
Source: MLEB data
Australian Tax Guidance Retrieval
Source: MLEB data
Australian Tax Guidance Retrieval
Bar Exam QA
Source: MLEB data
Consumer Contracts QA
Source: MLEB data
Consumer Contracts QA
Contractual Clause Retrieval
Source: MLEB data
Contractual Clause Retrieval
ECHR Retrieval
Source: MLEB data
Show 6 more resultsHide 6 results
GDPR Holdings Retrieval
Source: MLEB data
GDPR Holdings Retrieval
Irish Legislative Summaries
Source: MLEB data
Irish Legislative Summaries
License TL;DR Retrieval
Source: MLEB data
License TL, DR Retrieval
Singaporean Judicial Keywords
Source: MLEB data
Singaporean Judicial Keywords
UK Legislative Long Titles
Source: MLEB data
UK Legislative Long Titles

Vision

ViDoRe(v1)Source: MTEB Multimodal Boards data docs

Data credits: MLEB (Isaacus), MIT (licence)

Around the web

  • Jina Embeddings v4: Universal Embeddings for Multimodal Multilingual Retrieval
    Word "Embeddings" followed by a numeric or symbol representation, displayed in multiple colors on a …Word "Embeddings" followed by a numeric or symbol representation, displayed in multiple colors on a technology-themed, colorf Jina AI Jina AI · 12 minutes read jina-embeddings-v4 - Search Foundation …

    jina.ai

  • jina-embeddings-v5-omni: Embeddings for Text, Image, Audio and Video
    search reorder star Featured Press release May 12, 2026 jina-embeddings-v5-omni: Embeddings for Text, Image, Audio …search reorder star Featured Press release May 12, 2026 jina-embeddings-v5-omni: Embeddings for Text, Image, Audio and Video jina-embeddings-v5-omni-small is the best-performing open-weight omni embe…

    jina.ai

  • Jina Embeddings v3: A Frontier Multilingual Embedding Model
    jina-embeddings-v3 is a frontier multilingual text embedding model with 570M parameters and 8192 token-length, outperforming …jina-embeddings-v3 is a frontier multilingual text embedding model with 570M parameters and 8192 token-length, outperforming the latest proprietary embeddings from OpenAI and Cohere on MTEB. Dynamic …

    jina.ai

  • Elastic - Elastic Jina Embeddings v3 Now Available in Gemini Enterprise Agent Platform Model Garden
    April 23, 2026 High-performance, self-deployable multilingual embeddings deliver secure enterprise retrieval with full data control …April 23, 2026 High-performance, self-deployable multilingual embeddings deliver secure enterprise retrieval with full data control and predictable pricing SAN FRANCISCO--(BUSINESS WIRE)-- Elastic (N…

    ir.elastic.co

  • Jina Embeddings V3: Multilingual Embeddings With Task LoRA
    This paper introduces jina-embeddings-v3, a novel text embedding model with 570 million pa- rameters, optimized …This paper introduces jina-embeddings-v3, a novel text embedding model with 570 million pa- rameters, optimized for multilingual data, long- context retrieval, and high performance across multiple ta…

    arxiv.org

  • Jina Embeddings V3: Multilingual Embeddings With Task LoRA
    v3, a novel text embedding model with 570 million parameters, achieves state-of-the-art perfor- mance on …v3, a novel text embedding model with 570 million parameters, achieves state-of-the-art perfor- mance on multilingual data and long-context retrieval tasks, supporting context lengths of up to 8192 t…

    arxiv.org

  • jinaai/jina-embeddings-v5-text-nano · Hugging Face
    Notebooks Google Colab Kaggle jina-embeddings-v5-text-nano Model Overview Training and Evaluation Usage Optional / Recommended License …Notebooks Google Colab Kaggle jina-embeddings-v5-text-nano Model Overview Training and Evaluation Usage Optional / Recommended License Citation Jina AI: Your Search Foundation, Supercharged! jina-emb…

    huggingface.co

  • jinaai/jina-embeddings-v2-base-en · Hugging Face
    million parameters (you are here) . jina-embeddings-v2-base-zh : Chinese-English Bilingual embeddings. jina-embeddings-v2-base-de : German-English Bilingual …million parameters (you are here) . jina-embeddings-v2-base-zh : Chinese-English Bilingual embeddings. jina-embeddings-v2-base-de : German-English Bilingual embeddings. jina-embeddings-v2-base-es : S…

    huggingface.co

  • Better Text Embeddings: Jina AI launches jina-embeddings-v3, a text embedding model with task-specific adapters
    LoRA adapters, tailored for tasks like sentiment classification. Share * * * Text embedding models …LoRA adapters, tailored for tasks like sentiment classification. Share * * * Text embedding models are often used to retrieve text, cluster text, determine similarity between texts, and generate init…

    deeplearning.ai

Like this one. Tap or hover any dotted value, label or icon to see where it comes from and how it is measured. open details when tapped or hovered. How this data works