How to use from
llama.cppInstall from WinGet (Windows)
winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf FetchieAI/gemma-3-4b-lawft:BF16# Run inference directly in the terminal:
llama cli -hf FetchieAI/gemma-3-4b-lawft:BF16Build from source code
git clone https://github.com/ggml-org/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-app# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama serve -hf FetchieAI/gemma-3-4b-lawft:BF16# Run inference directly in the terminal:
./build/bin/llama cli -hf FetchieAI/gemma-3-4b-lawft:BF16Use Docker
docker model run hf.co/FetchieAI/gemma-3-4b-lawft:BF16Quick Links
TL;DR
This is a GGUF version of Gemma 3 4B, fine-tuned on Traditional Chinese legal instructions using court decisions published by the Judicial Yuan of Taiwan. It is suitable for legal Q&A, judgment summarization, and legal document reasoning. Compatible with llama.cpp and Ollama.
Training Data
This model was trained using a curated dataset of 96,556 Taiwanese court decisions from 2021, each with available full text (JFULL) and downloadable PDF (JPDF), where tabular structures are clearly embedded. The dataset is designed for:
- Table extraction
- OCR pretraining for legal PDFs
- Multimodal training (PDF + table reasoning)
Acknowledgments
Data collected from the publicly available Judicial Yuan open datasets (Taiwan).
- Downloads last month
- 10
Hardware compatibility
Log In to add your hardware
16-bit
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support
Install (macOS, Linux)
# Start a local OpenAI-compatible server with a web UI: llama serve -hf FetchieAI/gemma-3-4b-lawft:BF16# Run inference directly in the terminal: llama cli -hf FetchieAI/gemma-3-4b-lawft:BF16