Beijing Academy of Artificial Intelligence (BAAI)

BAAI’s LLM-based code embedding model for code and text retrieval. It supports natural-language queries in English and Chinese, covers 20 programming languages, and is also designed to keep strong multilingual text-retr…

BAAI’s multimodal retrieval family for image-text search. It includes lightweight CLIP-based models and larger MLLM-based variants, ranging from 150M to 7.57B parameters, and is built for matching images with text or mi…

• 🔍 Vision search
• 🔍 Image search

BAAI’s English embedding model with in-context learning. It is designed to improve retrieval on new tasks by adding a few-shot examples into the query, and BAAI positions it as a state-of-the-art BGE-series embedder on…

BAAI’s layer-selectable multilingual reranker in the v2 family. It is built to let users choose output layers for accelerated inference, trading some depth for speed while keeping strong multilingual reranking performan…

• 🔍 Information retrieval
• 🌐 Multilingual communication

BAAI’s larger multilingual reranker in the v2 family. It is designed for stronger multilingual reranking performance, with the docs highlighting good English proficiency alongside broad multilingual capability.

• 🔍 Information retrieval
• 🌐 Multilingual communication

BAAI’s lightweight multilingual reranker in the v2 family. It is built for fast inference and easier deployment while keeping strong multilingual reranking capability, making it the efficiency-oriented option in the v2…

• 🔍 Information retrieval
• 🌐 Multilingual communication

BAAI’s multilingual retrieval embedding model built around three strengths: multi-functionality, multi-linguality, and multi-granularity. It supports dense retrieval, sparse retrieval, and multi-vector retrieval in one…

• 🔍 Information retrieval
• 🌐 Multilingual communication
• 🔍 Conceptual search

BAAI’s lighter first-generation cross-encoder reranker for English and Chinese. It directly scores query-document relevance rather than producing embeddings, and is positioned as the faster, easier-to-deploy reranking o…

• 🔍 Information retrieval
• 📂 Content categorization

BAAI’s larger first-generation cross-encoder reranker for English and Chinese. Instead of producing embeddings, it scores a query-document pair directly, making it suitable for reranking top-k retrieval results with hig…

• 🔍 Information retrieval
• 📂 Content categorization

BAAI’s top English BGE v1.5 embedding model. It is the large-scale Sep-2023 update to bge-large-en, with the docs highlighting more reasonable similarity distribution and better performance than the original version.

• 🔍 Information retrieval
• 📂 Content categorization
• 🔍 Conceptual search

BAAI’s mid-size English BGE v1.5 embedding model. It belongs to the Sep-2023 v1.5 update that improved retrieval behavior and similarity distribution, while keeping a balanced base-scale footprint for practical deployme…

• 🔍 Information retrieval
• 📂 Content categorization
• 🔍 Conceptual search