CoolFace
Modelpublic

cstr/arctic-embed-m-v2-GGUF

sourceHugging Faceapache-2.0updated 2mo agoView on Hugging Face
0likes509downloads
Model Card

arctic-embed-m-v2 GGUF

GGUF format of Snowflake/snowflake-arctic-embed-m-v2.0 for use with CrispEmbed.

Snowflake Arctic Embed M v2.0. Multilingual retrieval embeddings, 768-dimensional, 8192-token context. Queries take the prefix "query: "; documents take no prefix.

Files

FileQuantizationSize
arctic-embed-m-v2-q4_k.ggufQ4_K261 MB
arctic-embed-m-v2-q8_0.ggufQ8_0315 MB
arctic-embed-m-v2.ggufF321170 MB

Quick Start

bash
# Download
huggingface-cli download cstr/arctic-embed-m-v2-GGUF arctic-embed-m-v2-q4_k.gguf --local-dir .

# Run with CrispEmbed
./crispembed -m arctic-embed-m-v2-q4_k.gguf "Hello world"

# Or with auto-download
./crispembed -m arctic-embed-m-v2 "Hello world"

Model Details

PropertyValue
ArchitectureGTE v1.5 (RoPE, GeGLU, post-LN)
Parameters305M
Embedding Dimension768
Layers12
PoolingCLS
TokenizerSentencePiece
Base ModelSnowflake/snowflake-arctic-embed-m-v2.0

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.

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

# Encode
./build/crispembed -m arctic-embed-m-v2-q4_k.gguf "query text"

# Server mode
./build/crispembed-server -m arctic-embed-m-v2-q4_k.gguf --port 8080
curl -X POST http://localhost:8080/v1/embeddings \
    -d '{"input": ["Hello world"], "model": "arctic-embed-m-v2"}'

Credits