Granite Embedding Small English R2
Granite Embedding Small English R2 is a 47M ModernBERT bi-encoder. It produces 384-dimensional vectors and succeeds IBM's earlier 30M English embedding release.
Model design and output
The smaller architecture uses fewer layers than the full English R2 encoder. It retains an 8,192-token input limit and encodes documents separately from queries. The result supports similarity search rather than natural-language generation.Granite Embedding Small English R2 model card
Practical use
Practical guidance. This is a candidate for a compact English search service. Compare missed relevant results with the larger English R2 model before choosing it for resource savings. A reranker can reorder retrieved items, but cannot recover a relevant item that the first-stage search omitted.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Launch date | |
| Access | Downloadable |
| Model identifiers | ibm-granite/granite-embedding-small-english-r2 |
| Context window | 8,192 tokens Published maximum input length. Granite Embedding Small English R2 model card |
| License | Apache 2.0Granite Embedding Small English R2 model card |
| Vector size | 384 dimensionsGranite Embedding Small English R2 model card |
Related models
- Granite Embedding Reranker English R2
IBM · Granite Reranker
A 149M ModernBERT cross-encoder that scores English query-document pairs to refine search results.
Launch date Downloadable - Granite Embedding English R2
IBM · Granite Embedding
A 149M English retrieval model that produces 768-dimensional vectors.
Launch date Downloadable