all-MiniLM-L6-v2

Permissive

Sentence-Transformers · MiniLM · Released 2021-08

22MEmbedding

The default lightweight embedding baseline for years: tiny, extremely fast, and good enough for most semantic search prototypes.

Strengths

  • +Apache 2.0 license
  • +Tiny (22M params), runs fast even on CPU
  • +Huge existing ecosystem and index compatibility

Limitations

  • -Retrieval quality trails larger 2023+ embedders like BGE-M3 or E5-Mistral
  • -256-token input limit is short for long-document chunks

License

Apache 2.0

use commercially with attribution niceties

Hardware

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

Links

Stats

sentence-transformers/all-MiniLM-L6-v2— downloads·— likes

via Hugging Face

Related models