TyKaoz/bge-m3-4bit

BGE-M3 — 4-bit MLX

4-bit MLX quantization of BAAI/bge-m3, for Apple Silicon (~321 MB). Multilingual text-embedding model (1024-dim dense vectors).

Usage

pip install -U mlx-embeddings

from mlx_embeddings import load, generate

model, tokenizer = load("TyKaoz/bge-m3-4bit")
out = generate(model, tokenizer, texts=["Bonjour", "Hello", "你好"])
print(out.text_embeds.shape)  # (3, 1024)

Base	Tool	Precision	Size
BAAI/bge-m3	mlx-embeddings	4-bit · group 64	~321 MB

By TyKaoz — privacy-first native macOS LLM chat client. Apache 2.0, inherited from the base model.