Granite Embedding 311M Multilingual R2
Granite Embedding 311M Multilingual R2 is a ModernBERT retrieval encoder with 768-dimensional output. It supports Matryoshka dimensions of 512, 384, 256 and 128 for smaller stored vectors.
Model design and output
The model uses a vocabulary of roughly 262,000 tokens. Its code training includes Python, Go, Java, JavaScript, PHP, Ruby, SQL, C and C++. The bi-encoder generates separate query and document vectors for similarity search.Granite Embedding 311M Multilingual R2 model card
Practical use
Practical guidance. Choose the vector dimension using retrieval quality and storage measurements together. Lower dimensions may suit a large collection, but they can lose distinctions that matter for close technical matches. Rebuild the index consistently when changing dimensions or model revisions.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | Downloadable |
| Model identifiers | ibm-granite/granite-embedding-311m-multilingual-r2 |
| Context window | 32,768 tokens Published maximum input length. Granite Embedding 311M Multilingual R2 model card |
| License | Apache 2.0Granite Embedding 311M Multilingual R2 model card |
| Parameters | 311MGranite Embedding 311M Multilingual R2 model card |
| Language coverage | Pretraining in 200+ languages; retrieval improvements for 52+ languages and code.Granite Embedding 311M Multilingual 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 97M Multilingual R2
IBM · Granite Embedding
A multilingual ModernBERT retrieval encoder that produces 384-dimensional text and code vectors.
Launch date not confirmedDownloadable