f2llm-v2-160m GGUF

GGUF format of codefuse-ai/F2LLM-v2-160M for use with CrispEmbed.

F2LLM Embedding v2 160M. Qwen3-based, pruned from the 0.6B base; the strongest sub-200M German embedder on MTEB(deu, v1).

Files

File	Quantization	Size
f2llm-v2-160m-q8_0.gguf	Q8_0	166 MB
f2llm-v2-160m.gguf	F32	494 MB

Parity vs HuggingFace reference

Cosine similarity vs the upstream sentence-transformers reference on a fixed test set (text):

Quant	Text
f16	1.0000
q8_0	0.9994

Quick Start

# Download
huggingface-cli download cstr/f2llm-v2-160m-GGUF f2llm-v2-160m-q8_0.gguf --local-dir .

# Run with CrispEmbed
./crispembed -m f2llm-v2-160m-q8_0.gguf "Hello world"

# Or with auto-download
./crispembed -m f2llm-v2-160m "Hello world"

Model Details

Property	Value
Architecture	Qwen3
Parameters	160M
Embedding Dimension	640
Layers	9
Pooling	last-token
Tokenizer	GPT-2 BPE
Base Model	codefuse-ai/F2LLM-v2-160M

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 f2llm-v2-160m-q8_0.gguf "query text"

# Server mode
./build/crispembed-server -m f2llm-v2-160m-q8_0.gguf --port 8080
curl -X POST http://localhost:8080/v1/embeddings \
    -d '{"input": ["Hello world"], "model": "f2llm-v2-160m"}'