Return to Articles. Enterprise Article, published on May 14, 2026. Radu Florian. – Ce-i asta? – Nu ştiu. Nu ştiu. Se numeşte IBM-Granit. Parul Awasthy (pawasthy) from the IBM Granite team. Aashka Trivedi. aashkaaa. ibm-granite. Madison Lee, sa tead, jõulupuu. Lisatud on järgmine tabel: **TL;DR:** Introducing two new Apache 2.0 multilingual embedding models based on ModernBERT — a compact 97M-parameter version that outperforms all open sub-100M multilingual embedders on MTEB Multilingual Retrieval (60.3), and a 311M full-size model that achieves 65.2 on the same benchmark (ranking #2 among open models under 500M parameters) while supporting Matryoshka embeddings. Both models support 200+ languages, are fine-tuned on 52 languages, feature a 31.33K-token context window (64× larger than R1), and include code retrieval capabilities across 9 programming languages.
In this post:
Enterprise-Ready by Design · A Strong Sub-100M Multilingual Model · What Changed from R1 · Training the Full-Size 311M Model · Building the Compact 97M Multilingual Model · Benchmark Results · Matryoshka Embeddings · Deployment Options · For Framework Integrators · Which Model Should You Use?
Hugging Face – Blog