CoolFace
Modelpublic

majentik/nomic-embed-text-v2-moe-GGUF-Q4_K_M

sourceHugging Faceapache-2.0updated 1mo agoView on Hugging Face
0likes51downloads
Model Card

nomic-embed-text-v2-moe GGUF Q4KM

llama.cpp GGUF Q4KM quantization of nomic-ai/nomic-embed-text-v2-moe (multilingual MoE text embedding model, 475M total / 305M active parameters).

  • —Source revision: 1066b6599d099fbb93dfcb64f9c37a7c9e503e85 (FP32 checkpoint)
  • —Converted via convert_hf_to_gguf.py (fresh llama.cpp tree), quantized with llama-quantize
  • —Quant type: Q4_K_M
  • —File size: 336 MB

Embedding fidelity gate

This tier passed the release gate: min cosine similarity 0.992074 (mean 0.993567) versus the F16 GGUF baseline across an 8-sentence multilingual set via llama-embedding (threshold 0.98).

Quickstart

bash
llama-embedding -m nomic-embed-text-v2-moe-Q4_K_M.gguf --pooling mean \
  -p "What is the capital of France?"

License

Apache 2.0 — inherited from the upstream base model.

See also