GemmaTranslate-v3-12B-GGUF

This repository contains the GGUF quantized versions of GemmaTranslate-v3-12B, a multimodal hybrid model merged from Gemma-3-12B-IT and TranslateGemma-12B-IT.

Available Formats

File	Size	Quantization	Description
GemmaTranslate-v3-BF16.gguf	~23.5GB	BF16	Full precision (source of truth).
GemmaTranslate-v3-Q8_0.gguf	~12.5GB	Q8_0	High fidelity, recommended for 16GB+ RAM.
GemmaTranslate-v3-Q4_K_M.gguf	~7.3GB	Q4_K_M	Balanced performance/size, recommended for 8GB+ RAM.

Usage with llama.cpp

./llama-cli -m GemmaTranslate-v3-Q4_K_M.gguf -p "Translate 'AI is the future' to French:"

Usage with Ollama

ollama run hf.co/SpongeBOB9684/GemmaTranslate-v3-12B-GGUF:Q4_K_M

License

This model is subject to the Gemma Terms of Use.