Jina Embeddings v4
Universal multimodal embedding model from Jina AI supporting text and images through unified pathway. Built on Qwen2.5-VL-3B-Instruct, outperforms proprietary models on visually rich document retrieval. This is a commercial API with free tier, though OSS weights available.
About this tool
Overview
jina-embeddings-v4 is a 3.8B parameter model that embeds text and images through a unified pathway, supporting both dense and late-interaction retrieval. Particularly strong on visually rich document retrieval, outperforming proprietary models from Google, OpenAI, and Voyage AI.
Key Features
- Multimodal: Unified embedding for text and images
- 3.8B Parameters: Built on Qwen2.5-VL-3B-Instruct foundation
- Dense + Late Interaction: Supports multiple retrieval modes
- 1,536 Dimensions: Compatible with many vector databases
- Open Weights: Available on Hugging Face for self-hosting
- API Access: Managed API with multiple tiers
Pricing
Token-Based Pricing
- Cost: Approximately $0.02 per million tokens
- Free Trial: 10 million tokens for new users with auto-generated API key
Rate Limits by Tier
- Free: 100 RPM, 100K TPM, 2 concurrent requests
- Paid: 500 RPM, 2M TPM, 50 concurrent requests
- Premium: 5,000 RPM, 50M TPM, 500 concurrent requests
Image Token Calculation
- Each tile costs 10 tokens
- Tiles are 28x28 pixels
- Image processing cost varies with image size
Pricing Model Update
New pricing model introduced May 6, 2025. Users with auto-recharge enabled before this date maintain old pricing. New pricing applies to new purchases or modifications.
Important Note on Throughput
Jina intentionally throttles API throughput for jina-embeddings-v4 to manage infrastructure costs. For production workloads requiring high throughput:
- Use jina-embeddings-v3 API, or
- Deploy jina-embeddings-v4 on your own infrastructure via Hugging Face
Payment Methods
Payments processed through Stripe supporting:
- Credit cards
- Google Pay
- PayPal
Model Access
- API: https://jina.ai/embeddings/
- Hugging Face: jinaai/jina-embeddings-v4
- Self-Hosting: Deploy on your infrastructure
- Cloud Marketplaces: Azure Marketplace
Use Cases
- Visually rich document retrieval
- Multimodal semantic search
- Document understanding with layout
- Cross-modal retrieval (text→image, image→text)
- RAG systems with visual content
Comparison to v3
v4 adds multimodal capabilities (text + images) with 1,536-dimensional vectors, while v3 was text-only with 1,024 dimensions. v3 offers higher API throughput for production text-only workloads.
Loading more......
Information
Categories
Tags
Similar Products
6 result(s)Commercial embedding models built for enterprise-grade semantic search and RAG applications. Features voyage-3 and voyage-3-large models with multimodal support. This is a commercial API service with usage-based pricing.
AI data lake with revolutionary index-on-the-lake technology enabling sub-second queries from S3. Features 10x cost efficiency vs in-memory DBs and 2x faster than alternatives. This is a commercial platform with OSS components.
Search and analytics engine with k-nearest neighbor (kNN) search for semantic similarity. Features approximate and exact kNN, HNSW indexing, and advanced quantization. This is commercial with OSS version available.
AI Search and RAG-as-a-Service platform with semantic search capabilities. Features NucliaDB open-source database. Acquired by Progress in 2025, now part of Progress Agentic RAG. This is a commercial service with OSS core (NucliaDB).
Open-source toolkit for developing AI applications using Postgres and pgvector. Provides managed PostgreSQL with built-in vector support, Python client (vecs), and AI features. This is a commercial managed service with OSS components.
An open-source library for creating, storing, and searching multimodal data and vector embeddings, supporting AI and ML workflows.