File	Quantization	Quality	Use Case
`qwen2.5-7b-sft-capybara-q4_k_m.gguf`	Q4_K_M	Good	Recommended - Best balance of quality and size
`qwen2.5-7b-sft-capybara-q5_k_m.gguf`	Q5_K_M	Better	Higher quality, moderate size increase
`qwen2.5-7b-sft-capybara-q8_0.gguf`	Q8_0	Best	Highest quality quantization

Download Specific Quantization

Using huggingface-cli

bash

# Download Q4_K_M (recommended)
huggingface-cli download ermiaazarkhalili/Qwen2.5-7B-SFT-Capybara-GGUF qwen2.5-7b-sft-capybara-q4_k_m.gguf --local-dir ./models

# Download Q5_K_M (higher quality)
huggingface-cli download ermiaazarkhalili/Qwen2.5-7B-SFT-Capybara-GGUF qwen2.5-7b-sft-capybara-q5_k_m.gguf --local-dir ./models

# Download Q8_0 (best quality)
huggingface-cli download ermiaazarkhalili/Qwen2.5-7B-SFT-Capybara-GGUF qwen2.5-7b-sft-capybara-q8_0.gguf --local-dir ./models

# Download all quantizations
huggingface-cli download ermiaazarkhalili/Qwen2.5-7B-SFT-Capybara-GGUF --local-dir ./models

Using wget

bash

# Q4_K_M
wget https://huggingface.co/ermiaazarkhalili/Qwen2.5-7B-SFT-Capybara-GGUF/resolve/main/qwen2.5-7b-sft-capybara-q4_k_m.gguf

# Q5_K_M
wget https://huggingface.co/ermiaazarkhalili/Qwen2.5-7B-SFT-Capybara-GGUF/resolve/main/qwen2.5-7b-sft-capybara-q5_k_m.gguf

# Q8_0
wget https://huggingface.co/ermiaazarkhalili/Qwen2.5-7B-SFT-Capybara-GGUF/resolve/main/qwen2.5-7b-sft-capybara-q8_0.gguf

Usage

Ollama

bash

# Pull specific quantization
ollama pull hf.co/ermiaazarkhalili/Qwen2.5-7B-SFT-Capybara-GGUF:Q4_K_M

# Or create from local file
cat > Modelfile << EOF
FROM ./qwen2.5-7b-sft-capybara-q4_k_m.gguf
EOF

ollama create qwen2.5-7b-sft-capybara -f Modelfile
ollama run qwen2.5-7b-sft-capybara

llama.cpp

bash

# Run with llama-cli
./llama-cli -m qwen2.5-7b-sft-capybara-q4_k_m.gguf -p "Your prompt here" -n 256

# Run as server
./llama-server -m qwen2.5-7b-sft-capybara-q4_k_m.gguf --host 0.0.0.0 --port 8080

llama-cpp-python

python

from llama_cpp import Llama

llm = Llama(
    model_path="qwen2.5-7b-sft-capybara-q4_k_m.gguf",
    n_ctx=2048,
    n_gpu_layers=-1  # Use all GPU layers
)

output = llm(
    "What is machine learning?",
    max_tokens=256,
    temperature=0.7,
)
print(output['choices'][0]['text'])

LM Studio

Download the desired GGUF file from this repository
Open LM Studio and navigate to the Models tab
Click "Add Model" and select the downloaded GGUF file
Load the model and start chatting

GPT4All

Download the Q4_K_M GGUF file
Open GPT4All and go to Settings > Models
Add the GGUF file path
Select the model and start using

Original Model

This is a quantized version of ermiaazarkhalili/Qwen2.5-7B-SFT-Capybara. See the original model card for:

Training details and methodology
Dataset information
Performance metrics
Full usage examples with Transformers

Conversion Details

Property	Value
Source Model	ermiaazarkhalili/Qwen2.5-7B-SFT-Capybara
Conversion Date	2026-04-09
Quantizations	Q4_K_M, Q5_K_M, Q8_0
Converter	llama.cpp

License

Same license as the original model. See ermiaazarkhalili/Qwen2.5-7B-SFT-Capybara for details.

Converted using the Slurm Model Trainer skill

Citation

If you use this model in your research or applications, please cite:

bibtex

@misc{azarkhalili2026_qwen2_5_7b_sft_capybara_gguf,
    author = {Azarkhalili, Behrooz},
    title = {Qwen2.5-7B-SFT-Capybara-GGUF},
    year = {2026},
    publisher = {Hugging Face},
    url = {https://huggingface.co/ermiaazarkhalili/Qwen2.5-7B-SFT-Capybara-GGUF}
}

To generate a citable DOI, click "Cite this model" on the model page.

Training Method

Trained using SFT (Supervised Fine-Tuning) on argilla/distilabel-capybara-dpo.

Base (merged) model: Qwen2.5-7B-SFT-Capybara

Acknowledgments

Hugging Face TRL Team for the training library
llama.cpp for the GGUF quantization format
Compute Canada / DRAC for HPC resources
Base model developers for making their weights openly available