BGE-M3

Permissive

BAAI · BGE · Released 2024-01

568MEmbedding

A versatile embedding model that unifies dense, sparse, and multi-vector (ColBERT-style) retrieval in one checkpoint across 100+ languages.

Strengths

  • +MIT license, unrestricted use
  • +Supports dense, sparse, and multi-vector retrieval from one model
  • +8192-token context, long for an embedding model

Limitations

  • -Multi-vector mode is slower and storage-heavier than plain dense embeddings
  • -Larger than lightweight embedders like all-MiniLM for simple use cases

License

MIT

use commercially with attribution niceties

Hardware

runs on a good consumer GPU (or Apple Silicon) with quantization

Links

Stats

BAAI/bge-m3— downloads·— likes

via Hugging Face

Related models