Embed 5 Pro
Embed 5 Pro is the quality-focused tier of Cohere's current embedding family. It represents text, images and mixed documents as vectors, and shares its vector space with Embed 5 Fast.
Model behavior
Outputs support float, int8 and binary formats. Cohere recommends Pro for indexing and difficult retrieval tasks. Its shared space permits Fast to encode queries against the same index. API and private deployment options are available; this is not a public weight release.Embed 5 release and technical overview
How to use it well
Practical guidance. Pro is a useful starting point for an archive whose documents are indexed less often than they are searched. Compare dimensions and storage formats on actual questions before choosing an index layout. Include visually complex pages and cases where a query uses different wording from the source. Preserve document identifiers and page locations beside the vectors so a retrieved match can become inspectable evidence.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | API |
| Model identifiers | embed-v5.0-pro |
| Context window | 128,000 tokens Maximum input context reported for Embed 5. Embed 5 release and technical overview |
| Model scale | Approximately 2.4 billion parametersCohere Embed specifications |
| Vector dimensions | 256, 512, 768, 1024, 1536 or 2048Embed 5 release and technical overview |
Related models
- Rerank 5 Pro
Cohere · Rerank
Quality-focused search reranker with Context Focus for returning relevant document passages.
Launch date not confirmedAPI - Embed 5 Fast
Cohere · Embed
Low-latency embedding model that shares its vector space with Embed 5 Pro.
Launch date not confirmedAPI Text, image and mixed-document embedding model with selectable vector dimensions and a 128K context.
Launch date not confirmedAPI