OpenEuroLLM-Czech-Ollama-GGUF

OpenEuroLLM-Czech (Ollama GGUF Mirror)

This repository mirrors the Ollama model jobautomation/OpenEuroLLM-Czech as a standalone GGUF plus the supporting Ollama metadata needed to recreate the package locally.

Contents

• OpenEuroLLM-Czech-Q4_K_M.gguf: the main GGUF weights file
• Modelfile: Ollama recipe adjusted to use the local GGUF file
• ollama-manifest.json: original Ollama manifest
• params.json: Ollama parameter block
• template.txt: Ollama chat template
• system.txt: default Czech system prompt
• licenses/: license materials shipped with the Ollama package
• NOTICE: required Gemma redistribution notice

Model Notes

• Architecture: gemma3
• Parameters: 12.2B
• Quantization: Q4_K_M
• Context length in GGUF metadata: 131072
• Default Ollama runtime context: 2048
• Primary language behavior: Czech

Use With Ollama

After downloading the repository locally:

ollama create OpenEuroLLM-Czech -f Modelfile

Use With GGUF Runtimes

The file OpenEuroLLM-Czech-Q4_K_M.gguf can also be used directly with GGUF-compatible runtimes such as llama.cpp and other tools that support Gemma 3 GGUF models.

Use With vLLM

This mirror now includes a root config.json and generation_config.json to improve compatibility with serving stacks that expect Hugging Face-style metadata.

For vLLM, you will typically still want to point the tokenizer at the original Gemma 3 tokenizer:

vllm serve vishinvents/OpenEuroLLM-Czech-Ollama-GGUF:Q4_K_M \
  --tokenizer google/gemma-3-12b-it

Licensing

This package includes the original license materials surfaced by Ollama:

• licenses/Gemma-Terms-of-Use.txt
• licenses/Additional-License.txt

Please review those files before use or redistribution.