cstr/snowflake-arctic-embed-m-GGUF

snowflake-arctic-embed-m GGUF

GGUF format of Snowflake/snowflake-arctic-embed-m for use with CrispEmbed.

Snowflake Arctic Embed M. Mid-range retrieval model, 768-dimensional CLS-pooled.

Files

File	Quantization	Size
snowflake-arctic-embed-m-q4_k.gguf	Q4_K	71 MB
snowflake-arctic-embed-m-q8_0.gguf	Q8_0	112 MB
snowflake-arctic-embed-m.gguf	F32	418 MB

Quick Start

# Download
huggingface-cli download cstr/snowflake-arctic-embed-m-GGUF snowflake-arctic-embed-m-q4_k.gguf --local-dir .

# Run with CrispEmbed
./crispembed -m snowflake-arctic-embed-m-q4_k.gguf "Hello world"

# Or with auto-download
./crispembed -m snowflake-arctic-embed-m "Hello world"

Model Details

Property	Value
Architecture	BERT
Parameters	109M
Embedding Dimension	768
Layers	12
Pooling	CLS
Tokenizer	WordPiece
Base Model	Snowflake/snowflake-arctic-embed-m

Verification

Verified bit-identical to HuggingFace sentence-transformers (cosine similarity >= 0.999 on test texts).

Usage with CrispEmbed

CrispEmbed is a lightweight C/C++ text embedding inference engine using ggml. No Python runtime, no ONNX. Supports BERT, XLM-R, Qwen3, and Gemma3 architectures.

# Build CrispEmbed
git clone https://github.com/CrispStrobe/CrispEmbed
cd CrispEmbed
cmake -S . -B build && cmake --build build -j

# Encode
./build/crispembed -m snowflake-arctic-embed-m-q4_k.gguf "query text"

# Server mode
./build/crispembed-server -m snowflake-arctic-embed-m-q4_k.gguf --port 8080
curl -X POST http://localhost:8080/v1/embeddings \
    -d '{"input": ["Hello world"], "model": "snowflake-arctic-embed-m"}'

Credits

• Original model: Snowflake/snowflake-arctic-embed-m
• Inference engine: CrispEmbed (ggml-based)
• Conversion: convert-bert-embed-to-gguf.py