Skip to content

BGE-M3

Hugging Face (BAAI) Open weights

The catch: Released January 2024 with no newer M3 version on BAAI's Hugging Face page; the card's sparse and multi-vector examples use BAAI's FlagEmbedding library.

AccessOpen weights
Free limitsFree to download and self-host; you pay only for your own compute. 1024-dim dense embeddings plus sparse and multi-vector (ColBERT) outputs from one model; inputs up to 8192 tokens; 100+ languages. The BAAI card states no parameter count (Ollama lists it as bge-m3:567m).
Modalityembeddings
Credit cardNot required
Commercial useAllowed
Trains on your data–
Licence or termsMIT
Context window8,192 tokens
Model idBAAI/bge-m3
Base URL–
Last checked2 Oct 2026

How to use BGE-M3

Download the weights and serve them yourself; the quickstart below shows one way. Licence: MIT.

Quickstart
ollama pull bge-m3
Paste into Claude Code
I want to run the open-weight model "BGE-M3" (BAAI/bge-m3) for my project.
Please:
1. Check the model card and tell me the hardware it needs and the simplest way to serve it (vLLM, SGLang, Ollama or a hosted endpoint).
2. Give me a minimal working setup and a test request.
3. Tell me the licence terms for commercial use.
4. If a cheaper or smaller variant would do the job, mention it.

Before you ship on it

  • Released January 2024 with no newer M3 version on BAAI's Hugging Face page; the card's sparse and multi-vector examples use BAAI's FlagEmbedding library.
  • MIT licence (licence tag on huggingface.co/BAAI/bge-m3); commercial use allowed.
  • Free to run means you pay for the GPUs, the serving stack and the on-call.

Questions

Is BGE-M3 free?

Yes. Open weights offers it as open weights, with these limits: Self-host free · MIT. No credit card is needed.

Can I use BGE-M3 commercially?

Yes, under MIT. Check the current terms before you ship.

How do I start with BGE-M3?

Download the weights and serve them yourself; the quickstart below shows one way. Licence: MIT.

Source: huggingface.co/BAAI/bge-m3, checked 2 Oct 2026.