Embed 5 Pro
Available nowCohere’s higher-capability embedding option for demanding multilingual enterprise retrieval.
What it does
Cohere positions Embed 5 Pro as its most capable embedding model for text and images, with support for more than 100 languages. It represents information for semantic search, classification and clustering, helping teams find related content without relying on exact words.
The official model guide describes a 128,000-token source context and several output dimensions. embeddings.app integrates text retrieval and explicitly requests 1,536 dimensions by default; this differs from the provider’s 2,048-dimensional default.
Language coverage
Cohere documents multilingual enterprise retrieval across more than 100 languages. Coverage does not mean equal performance across every language pair.
Official language coverage ↗Content types
text · image
Enterprise retrieval over multilingual text, charts and document images. Embed PDF page images rather than raw PDF files.
AIRuntime media adapter implemented; commercial media activation is pending. The media API awaits commercial activation.
Where it helps
Enterprise knowledge
Retrieve relevant text from complex multilingual collections.
Service assistants
Find source material for detailed customer questions.
Content organization
Build similarity-based discovery over related business documents.
When to choose it
A candidate for a new Cohere retrieval project where quality is the priority. Compare actual results with an established model and account for index storage as well as the cost of new inputs.
The AIRuntime image-and-text adapter and media API are implemented, with separate commercial media activation pending. Cohere takes images of PDF pages, not raw PDFs. The provider context is larger than this service’s input limit.
Official sources
Paraphrased from the provider’s documentation and original model information. Selection advice is our editorial guidance.
Cohere · Embed models ↗Cohere · Embeddings guide ↗Reviewed October 3, 2026