bge-base-en

Maintained By
BAAI

BGE Base English Embedding Model

PropertyValue
Parameter Count109M parameters
LicenseMIT
PaperC-Pack: Packaged Resources To Advance General Chinese Embedding
ArchitectureBERT-based Transformer

What is bge-base-en?

BGE-base-en is a powerful embedding model developed by BAAI that creates high-quality text embeddings for English language content. It's specifically designed to transform text into dense vector representations that are optimized for similarity search, retrieval, and various natural language processing tasks. With 109M parameters, it offers an excellent balance between performance and computational efficiency.

Implementation Details

The model architecture is based on BERT and has been trained using both RetroMAE pre-training and contrastive learning on large-scale paired data. It generates 768-dimensional embeddings and supports a maximum sequence length of 512 tokens.

  • Achieves state-of-the-art performance on the MTEB benchmark
  • Supports efficient inference with both PyTorch and ONNX
  • Includes specialized query instruction handling for retrieval tasks

Core Capabilities

  • Text Embeddings Generation: Creates high-quality vector representations for text similarity tasks
  • Semantic Search: Optimized for retrieval tasks with strong performance on benchmark datasets
  • Cross-encoder Compatibility: Can be used alongside BGE reranker models for enhanced accuracy
  • Multi-framework Support: Compatible with popular frameworks like Sentence-Transformers and Hugging Face Transformers

Frequently Asked Questions

Q: What makes this model unique?

The model achieves exceptional performance on the MTEB benchmark while maintaining a moderate parameter count, making it particularly efficient for production deployments. It also features specialized handling of query instructions for improved retrieval performance.

Q: What are the recommended use cases?

The model excels in semantic search, document retrieval, text similarity comparison, and clustering applications. It's particularly well-suited for production environments where a balance of performance and efficiency is crucial.

The first platform built for prompt engineering