Embed v4 — reviews, specs & pricing
Cohere's multimodal embedding model for text and images.
Summary
Embed v4 generates unified embeddings for text, images and mixed documents (like PDFs with charts), supporting a 128k token context. It's designed for enterprise search and RAG over multimodal content. It supports multiple languages.
Sample use case
Used for multimodal semantic search, document retrieval, and enterprise knowledge bases. Good for RAG systems ingesting mixed media content.
Specifications
- Provider: cohere
- License: closed
- Context: 128k tokens
- Input price: $0.12/M tok
- Released: 2025-04-15
Pros
- Multimodal embeddings
- Long context
- Multilingual
Cons
- Closed weights
- Cost scales with volume
- Newer, fewer integrations
Average rating 0.0 from 0 community reviews on Reviuws.