README.md

metadata

license: cc-by-nc-4.0
tags:
  - vidore
  - colpali
  - multimodal-embedding
  - multilingual-embedding
  - Text-to-Visual Document (T→VD) retrieval
  - feature-extraction
  - sentence-similarity
  - mteb
  - sentence-transformers
language:
  - multilingual
inference: false
library_name: transformers
pipeline_tag: visual-document-retrieval

The embedding model trained by Jina AI.

Jina Embeddings v4: Universal Embeddings for Multimodal Multilingual Retrieval

Quick Start

Blog | Technical Report | API

Intended Usage & Model Info

jina-embeddings-v4 is a universal embedding model for multimodal and multilingual retrieval. The model is specially designed for complex document retrieval, including visually rich documents with charts, tables, and illustrations.

Built on Qwen/Qwen2.5-VL-3B-Instruct, jina-embeddings-v4 features:

• Unified embeddings for text, images, and visual documents, supporting both dense (single-vector) and late-interaction (multi-vector) retrieval.
• Multilingual support (30+ languages) and compatibility with a wide range of domains, including technical and visually complex documents.
• Task-specific adapters for retrieval, text matching, and code-related tasks, which can be selected at inference time.
• Flexible embedding size: dense embeddings are 2048 dimensions by default but can be truncated to as low as 128 with minimal performance loss.

Summary of features:

Feature	Jina Embeddings V4
Base Model	Qwen2.5-VL-3B-Instruct
Supported Tasks	retrieval, text-matching, code
Model DType	BFloat 16
Max Sequence Length	32768
Single-Vector Dimension	2048
Multi-Vector Dimension	128
Matryoshka dimensions	128, 256, 512, 1024, 2048
Pooling Strategy	Mean pooling
Attention Mechanism	FlashAttention2

Training & Evaluation

Please refer to our technical report of jina-embeddings-v4 for training details and benchmarks.

Usage

Requirements

The following Python packages are required:

• transformers>=4.52.0
• torch>=2.6.0
• peft>=0.15.2
• torchvision
• pillow

Optional / Recommended

• flash-attention: Installing flash-attention is recommended for improved inference speed and efficiency, but not mandatory.
• sentence-transformers: If you want to use the model via the sentence-transformers interface, install this package as well.

via Jina AI Embeddings API

curl https://api.jina.ai/v1/embeddings \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $JINA_AI_API_TOKEN" \
  -d @- <<EOFEOF
  {
    "model": "jina-embeddings-v4",
    "task": "text-matching",
    "input": [
        {
            "text": "غروب جميل على الشاطئ"
        },
        {
            "text": "海滩上美丽的日落"
        },
        {
            "text": "A beautiful sunset over the beach"
        },
        {
            "text": "Un beau coucher de soleil sur la plage"
        },
        {
            "text": "Ein wunderschöner Sonnenuntergang am Strand"
        },
        {
            "text": "Ένα όμορφο ηλιοβασίλεμα πάνω από την παραλία"
        },
        {
            "text": "समुद्र तट पर एक खूबसूरत सूर्यास्त"
        },
        {
            "text": "Un bellissimo tramonto sulla spiaggia"
        },
        {
            "text": "浜辺に沈む美しい夕日"
        },
        {
            "text": "해변 위로 아름다운 일몰"
        },
        {
            "image": "https://i.ibb.co/nQNGqL0/beach1.jpg"
        },
        {
            "image": "https://i.ibb.co/r5w8hG8/beach2.jpg"
        }
    ]
  }
EOFEOF

via transformers

# !pip install transformers>=4.52.0 torch>=2.6.0 peft>=0.15.2 torchvision pillow
# !pip install
from transformers import AutoModel
import torch

# Initialize the model
model = AutoModel.from_pretrained("jinaai/jina-embeddings-v4", trust_remote_code=True)

model.to("cuda")

# ========================
# 1. Retrieval Task
# ========================
# Configure truncate_dim, max_length (for texts), max_pixels (for images), vector_type, batch_size in the encode function if needed

# Encode query
query_embeddings = model.encode_text(
    texts=["Overview of climate change impacts on coastal cities"],
    task="retrieval",
    prompt_name="query",
)

# Encode passage (text)
passage_embeddings = model.encode_text(
    texts=[
        "Climate change has led to rising sea levels, increased frequency of extreme weather events..."
    ],
    task="retrieval",
    prompt_name="passage",
)

# Encode image/document
image_embeddings = model.encode_image(
    images=["https://i.ibb.co/nQNGqL0/beach1.jpg"],
    task="retrieval",
)

# ========================
# 2. Text Matching Task
# ========================
texts = [
    "غروب جميل على الشاطئ",  # Arabic
    "海滩上美丽的日落",  # Chinese
    "Un beau coucher de soleil sur la plage",  # French
    "Ein wunderschöner Sonnenuntergang am Strand",  # German
    "Ένα όμορφο ηλιοβασίλεμα πάνω από την παραλία",  # Greek
    "समुद्र तट पर एक खूबसूरत सूर्यास्त",  # Hindi
    "Un bellissimo tramonto sulla spiaggia",  # Italian
    "浜辺に沈む美しい夕日",  # Japanese
    "해변 위로 아름다운 일몰",  # Korean
]

text_embeddings = model.encode_text(texts=texts, task="text-matching")

# ========================
# 3. Code Understanding Task
# ========================

# Encode query
query_embedding = model.encode_text(
    texts=["Find a function that prints a greeting message to the console"],
    task="code",
    prompt_name="query",
)

# Encode code
code_embeddings = model.encode_text(
    texts=["def hello_world():\n    print('Hello, World!')"],
    task="code",
    prompt_name="passage",
)

# ========================
# 4. Use multivectors
# ========================

multivector_embeddings = model.encode_text(
    texts=texts,
    task="retrieval",
    prompt_name="query",
    return_multivector=True,
)

images = ["https://i.ibb.co/nQNGqL0/beach1.jpg", "https://i.ibb.co/r5w8hG8/beach2.jpg"]
multivector_image_embeddings = model.encode_image(
    images=images,
    task="retrieval",
    return_multivector=True,
)

via sentence-transformers

from sentence_transformers import SentenceTransformer

# Initialize the model
model = SentenceTransformer("jinaai/jina-embeddings-v4", trust_remote_code=True)
# ========================
# 1. Retrieval Task
# ========================
# Encode query
query_embeddings = model.encode(
    sentences=["Overview of climate change impacts on coastal cities"],
    task="retrieval",
    prompt_name="query",
)

print(f"query_embeddings.shape = {query_embeddings.shape}")

# Encode passage (text)
passage_embeddings = model.encode(
    sentences=[
        "Climate change has led to rising sea levels, increased frequency of extreme weather events..."
    ],
    task="retrieval",
    prompt_name="passage",
)

print(f"passage_embeddings.shape = {passage_embeddings.shape}")

# Encode image/document
image_embeddings = model.encode(
    sentences=["https://i.ibb.co/nQNGqL0/beach1.jpg"],
    task="retrieval",
)

print(f"image_embeddings.shape = {image_embeddings.shape}")

# ========================
# 2. Text Matching Task
# ========================
texts = [
    "غروب جميل على الشاطئ",  # Arabic
    "海滩上美丽的日落",  # Chinese
    "Un beau coucher de soleil sur la plage",  # French
    "Ein wunderschöner Sonnenuntergang am Strand",  # German
    "Ένα όμορφο ηλιοβασίλεμα πάνω από την παραλία",  # Greek
    "समुद्र तट पर एक खूबसूरत सूर्यास्त",  # Hindi
    "Un bellissimo tramonto sulla spiaggia",  # Italian
    "浜辺に沈む美しい夕日",  # Japanese
    "해변 위로 아름다운 일몰",  # Korean
]

text_embeddings = model.encode(sentences=texts, task="text-matching")

# ========================
# 3. Code Understanding Task
# ========================

# Encode query
query_embeddings = model.encode(
    sentences=["Find a function that prints a greeting message to the console"],
    task="code",
    prompt_name="query",
)

# Encode code
code_embeddings = model.encode(
    sentences=["def hello_world():\n    print('Hello, World!')"],
    task="code",
    prompt_name="passage",
)

# ========================
# 4. Use multivectors
# ========================
# If you want to use multi-vector embeddings, please use the Hugging Face model directly.

Jina-VDR

Alongside jina-embeddings-v4, we’re releasing Jina VDR, a multilingual, multi-domain benchmark for visual document retrieval. The task collection can be viewed here, and evaluation instructions can be found here.

License

This model is licensed to download and run under CC BY-NC 4.0. It is available for commercial use via the Jina Embeddings API, AWS, Azure, and GCP. To download for commercial use, please contact us.

Contact

Join our Discord community and chat with other community members about ideas.

Citation

If you find jina-embeddings-v4 useful in your research, please cite the following paper:

@misc{günther2025jinaembeddingsv4universalembeddingsmultimodal,
      title={jina-embeddings-v4: Universal Embeddings for Multimodal Multilingual Retrieval}, 
      author={Michael Günther and Saba Sturua and Mohammad Kalim Akram and Isabelle Mohr and Andrei Ungureanu and Sedigheh Eslami and Scott Martens and Bo Wang and Nan Wang and Han Xiao},
      year={2025},
      eprint={2506.18902},
      archivePrefix={arXiv},
      primaryClass={cs.AI},
      url={https://arxiv.org/abs/2506.18902}, 
}