Modelpublic

applied-ai-subscr/ministral_3_3B_sudoku_gguf

sourceHugging Faceapache-2.0updated 10mo agoView on Hugging Face

0likes137downloads

Model Card

Ministral 3 3B Sudoku - GGUF

Quantized GGUF versions of the fine-tuned Ministral-3-3B model for Sudoku tasks.

Available Quantizations

—F16 (6.4 GB): 16-bit float, original quality
—Q8_0 (3.5 GB): 8-bit quantization, very good quality

Usage with llama.cpp

bash

# Download a model
huggingface-cli download applied-ai-subscr/ministral_3_3B_sudoku_gguf ministral-3-3b-sudoku-q8_0.gguf --local-dir ./models

# Run with llama.cpp
./llama-cli \
  -m ./models/ministral-3-3b-sudoku-q8_0.gguf \
  -c 4096 \
  -ngl 99 \
  -p "Solve this Sudoku..."

# Or start a server
./llama-server \
  -m ./models/ministral-3-3b-sudoku-q8_0.gguf \
  -c 4096 \
  -ngl 99 \
  --port 8080

Model Details

—Base model: unsloth/Ministral-3-3B-Instruct-2512
—Fine-tuned with Unsloth
—Converted to GGUF using llama.cpp converter