cstr/all-MiniLM-L6-v2-GGUF

all-MiniLM-L6-v2 GGUF

GGUF format of sentence-transformers/all-MiniLM-L6-v2 for use with CrispEmbed and Ollama.

Files

File	Quantization	Size
all-MiniLM-L6-v2-f32.gguf	F32	0 MB
all-MiniLM-L6-v2-q4_k.gguf	Q4_K	0 MB
all-MiniLM-L6-v2-q8_0.gguf	Q8_0	0 MB
all-MiniLM-L6-v2.gguf	F32	0 MB

Recommended: Q8_0 for quality (cos vs HF: 0.9998), Q4_K for size (0.970).

Quick Start

CrispEmbed

./crispembed -m all-MiniLM-L6-v2 "Hello world"
./crispembed-server -m all-MiniLM-L6-v2 --port 8080

Ollama (with CrispStrobe fork)

# Create model
echo "FROM all-MiniLM-L6-v2-q8_0.gguf" > Modelfile
ollama create all-MiniLM-L6-v2 -f Modelfile

# Embed
curl http://localhost:11434/api/embed -d '{"model":"all-MiniLM-L6-v2","input":["Hello world"]}'

Python (CrispEmbed)

from crispembed import CrispEmbed
model = CrispEmbed("all-MiniLM-L6-v2-q8_0.gguf")
vectors = model.encode(["Hello world", "Goodbye world"])

Model Details

Property	Value
Architecture	BERT
Parameters	22M
Embedding Dimension	384
Layers	6
Pooling	mean
Tokenizer	WordPiece
Language	en
Q8_0 vs HuggingFace	0.9998
Q4_K vs HuggingFace	0.970

Server API

CrispEmbed server supports four API dialects:

• POST /embed — native
• POST /v1/embeddings — OpenAI-compatible
• POST /api/embed — Ollama-compatible
• POST /api/embeddings — Ollama legacy

Credits

• Original model: sentence-transformers/all-MiniLM-L6-v2
• Inference: CrispEmbed (MIT, ggml-based)