AI Potluck
Contribute
Back to Gap Map Model components / Embeddings & retrieval models

BGE

Beijing Academy of Artificial Intelligence (BAAI)
open weights / Overall score: 4.3(strong)

BGE (BAAI General Embedding) covers text embedding and reranking for retrieval pipelines. The bge-m3 checkpoint works across more than 100 languages and inputs up to 8,192 tokens, emitting dense, sparse and multi-vector representations from a single pass; the English v1.5 encoders span small to large, and the bge-reranker line supplies cross-encoder relevance scores. The Beijing Academy of Artificial Intelligence maintains the family and the FlagEmbedding toolkit around it.

The family ships under three different licenses, and the two Gemma-based checkpoints are the ones that govern the openness score. Verified 2026-09-11 via the Hugging Face model cards, the Hub API and the FlagEmbedding repository README.

Openness

3 high confidence
3.0
weights
open(ungated safetensors and PyTorch checkpoints across the v1.5 encoders and bge-m3)
data
open(the BAAI-MTP pretraining corpus, the bge-m3-data fine-tuning set and the MLDR long-document set are all published)
code
open(FlagEmbedding ships embedder and reranker fine-tuning, unified fine-tuning, evaluation and inference pipelines under MIT)
license
MIT(OSI

Data and pipeline are as open as this ladder asks for, and the license is what caps the score. Most of the family is MIT or Apache-2.0, but two distributed checkpoints inherit the Gemma terms from Gemma 2, and the multi-SKU rule resolves on the most restrictive license whose weights actually ship.

Adoption

5 high confidence
5.0

139,255,026 downloads in the trailing 30 days across the bge EMBEDDING checkpoints in the BAAI account - 64,850,135 for bge-small-en-v1.5 alone and 37,891,167 for bge-m3. The reranker checkpoints are deliberately excluded: they draw a further 24,937,113 and are scored separately as `bge-reranker`, so counting all 24 bge-prefixed repos together, as an earlier reading of this record did, would bill the same downloads to two products. The band is 5 either way.

Capability

4 high confidence
4.0

Competitive frontier rather than leader: multilingual, long-context, and the best MIRACL figure in the comparison that ranks it, but its BEIR average trails the newer encoders and the family has not shipped a successor to m3.

Verified 2026-09-11