BGE-large-en-v1.5

BGE-large-en-v1.5

BGE-large-en-v1.5 (Beijing Academy of AI General Embedding) is a state-of-the-art English text embedding model released by BAAI under the MIT license. It consistently ranks at the top of the MTEB (Massive Text Embedding Benchmark) leaderboard for retrieval, reranking, and semantic similarity tasks.

Key features

• 1024-dimensional embeddings — high-fidelity semantic representation
• Top MTEB scores across retrieval, classification, and clustering tasks
• Dual-encoder architecture optimized for retrieval
• MIT license — fully commercial-friendly
• Works out-of-the-box with sentence-transformers, LangChain, and LlamaIndex
• Prefix instructions (Represent this sentence for retrieval:) boost retrieval performance

Quick start

from sentence_transformers import SentenceTransformer

model = SentenceTransformer("BAAI/bge-large-en-v1.5")

queries = ["Represent this sentence for retrieval: What is RAG?"]
docs = ["Retrieval-Augmented Generation grounds LLMs with external knowledge."]

q_emb = model.encode(queries, normalize_embeddings=True)
d_emb = model.encode(docs, normalize_embeddings=True)
scores = q_emb @ d_emb.T
print(scores)  # cosine similarity

Install via ai-supply

npx ai-supply add bge-large-en-v1-5

Curated mirror of the open-source BGE-large-en-v1.5 (MIT). Get it from the source.