Embed 5 Pro
AI model · Embedding model · Text → Embeddings by Cohere
The new industry standard for enterprise embeddings and maximum retrieval quality.
Embed 5 Pro is Cohere's state-of-the-art embedding model for enterprise retrieval, supporting multimodal inputs and 100+ languages. It is the ideal choice for businesses needing high-precision semantic search across complex documents and financial data.
Model specs
- Provider
- Cohere
- Model family
- Embed
- Version
- 5 Pro
- Released
- 2026-09-30
- Context window
- 128,000 tokens
- Input
- Text, Image
- Output
- Embeddings
- Capabilities
- Vision, Structured output
- Available on
- Cohere API, Amazon SageMaker, OpenRouter
- Launched
- 2026-09-30
- Languages
- English, French, Spanish, German, Chinese, Japanese, Arabic
Category: Natural Language Processing
- Modalities
- Text, Images, Fused inputs -> Vector
- Context window
- 128k tokens
- Sizes and variants
- Embed 5 Pro, Embed 5 Fast
- Released
- 2026-09-30
- Languages
- 100+ languages
- Input Modalities
- Text, Images, Fused (e.g., PDF)
- 128k token context window for long documents
- Supports text, images, and fused inputs like PDFs
- Multilingual support for over 100 languages
- Shared embedding space with Embed 5 Fast variant
Overview
Embed 5 Pro is Cohere's premier embedding model designed for high-quality enterprise retrieval and semantic search. It solves the problem of searching through fragmented and complex data by converting text, images, and fused inputs like PDFs into numerical vectors. The model is built for businesses requiring the highest retrieval accuracy across visually rich documents and financial filings. It supports a 128k token context window and over 100 languages, making it suitable for global search and RAG pipelines. Embed 5 Pro is available via the Cohere API, Model Vault, and major cloud platforms including Amazon SageMaker and Oracle Cloud Infrastructure.
Search and Retrieval
- State-of-the-Art Retrieval: Embed 5 Pro is optimized for the highest retrieval quality across visually rich documents, financial filings, and multimodal enterprise data.
- Multimodal Support: The model generates embeddings from text, images, and fused inputs such as PDF pages to unify visual and text data.
- Multilingual Reach: Embed 5 Pro supports over 100 languages for global search, classification, and clustering applications.
- Shared Embedding Space: Pro and Fast variants share an embedding space, allowing a corpus indexed with Pro to be queried with the Fast model.
Data Intelligence
- Semantic Understanding: The model captures the semantic meaning of text and images for accurate document comparison and pattern discovery.
- Efficient Vectorization: Embed 5 Pro converts complex enterprise data into numerical vectors for fast and scalable search operations.
- Enterprise Performance: The model is designed for quality-critical retrieval tasks, including offline indexing of large document corpora.
Use cases
- Investment teams use Embed 5 Pro to index and search through complex financial filings and visually rich reports.
- Developers build RAG pipelines that unify text and image data for more accurate AI agent responses.
- Global enterprises implement multilingual semantic search to find relevant documents across 100+ different languages.
- Data scientists use the model for high-quality document classification and clustering of large unstructured datasets.
Specifications
- Modalities
- Text, Images, Fused inputs -> Vector
- Context window
- 128k tokens
- Sizes and variants
- Embed 5 Pro, Embed 5 Fast
- Access
- Cohere API, Model Vault, OpenRouter, Amazon SageMaker, Oracle Cloud Infrastructure (OCI)
- Released
- 2026-09-30
- Languages
- 100+ languages
- Input Modalities
- Text, Images, Fused (e.g., PDF)
- Language Support
- 100+ languages
- Primary Endpoint
- /embed
- Deployment Options
- API, Model Vault, Private Cloud, On-Premise
How it works
- Get an API Key: Obtain a Cohere API key from the dashboard to begin integrating the model into applications.
- Install SDKs: Install the Cohere SDK for Python, TypeScript, Java, or Go to interact with the Embed endpoint.
- Index Your Data: Use the Embed 5 Pro model to convert your document corpus into numerical vectors for indexing.
- Query and Retrieve: Perform semantic searches or RAG queries against the indexed vectors to retrieve relevant information.
FAQ
What is the difference between Embed 5 Pro and Fast?
Embed 5 Pro is optimized for the highest retrieval quality and offline indexing, while Embed 5 Fast is built for low latency and high throughput in interactive search and agent loops.
Can I use Embed 5 Pro and Fast together?
Yes, they share an embedding space, meaning you can index your documents with the Pro model and perform queries using the Fast model.
What is the context window for Embed 5 Pro?
Embed 5 Pro supports a context window of 128,000 tokens.
What modalities does Embed 5 Pro support?
The model supports text, images, and mixed inputs such as PDF pages.
Added by AntSpace from public sources (cohere.com).