Granite Embedding R2 Models

2025-08-26

AI Summary

To address the challenge of balancing performance and efficiency for enterprise-scale dense retrieval across heterogeneous domains—including text, code, long documents, multi-turn dialogues, and tables—this paper introduces a high-performance English encoder embedding model family. Methodologically, it employs a dual-encoder–cross-encoder collaborative architecture with 22-layer and 12-layer variants, trained exclusively on compliant enterprise data; supports 8,192-token contexts (16× context expansion) and incorporates inference-path optimizations to jointly maximize accuracy and latency efficiency. Contributions include consistent state-of-the-art results across standard benchmarks, IBM’s internal evaluation suites, and production deployment scenarios; open-sourcing under the Apache 2.0 license; and delivering three specialized models—efficient dual-encoders, cross-encoders, and re-rankers—that significantly advance long-context understanding and multimodal content representation capabilities.

Technology Category

• Natural Language Processing: (Large) Language Models
• Machine Learning: Large Multimodal Models (LMMs)
• Computer Vision: Large Vision Models

Application Category

• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Web Mining and Content Analysis: Large pretrained models with web data
• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs

Abstract

We introduce the Granite Embedding R2 models, a comprehensive family of high-performance English encoder-based embedding models engineered for enterprise-scale dense retrieval applications. Building upon our first-generation release, these models deliver substantial improvements, including 16x expanded context length (8,192 tokens), state-of-the-art performance across diverse retrieval domains - text, code, long-document search, multi-turn conversational, and tabular data - and measurable speed advantages of 19-44% over leading competitors while maintaining superior accuracy. Our release encompasses both bi-encoder and cross-encoder architectures, featuring a highly effective 22-layer retriever model and its efficient 12-layer counterpart, alongside a high-quality reranker model, all trained exclusively on enterprise-appropriate data with comprehensive governance oversight. The models demonstrate exceptional versatility across standard benchmarks, IBM-developed evaluation suites, and real-world enterprise use cases, establishing new performance standards for open-source embedding models. In an era where retrieval speed and accuracy are paramount for competitive advantage, the Granite R2 models deliver a compelling combination of cutting-edge performance, enterprise-ready licensing, and transparent data provenance that organizations require for mission-critical deployments. All models are publicly available under the Apache 2.0 license at https://huggingface.co/collections/ibm-granite, enabling unrestricted research and commercial use.

Problem

• Develop high-performance embedding models for enterprise retrieval
• Achieve state-of-the-art accuracy with significant speed improvements
• Provide versatile retrieval capabilities across diverse data types

Innovation

• Expanded context length to 8192 tokens
• Combined bi-encoder and cross-encoder architectures
• Enterprise data training with governance oversight