---
base_model: google/gemma-4-12B-it
library_name: gguf
pipeline_tag: text-generation
tags:
- gemma4
- gguf
- llama.cpp
- q4_k_m
- quantized
- benign-refusal-reduction
---

# Gemma 4 12B IT Libre Q4_K_M GGUF

This repository contains a `Q4_K_M` GGUF build derived from `google/gemma-4-12B-it` plus a benign false-refusal reduction adapter merged into the text model.

It is not presented as an "abliterated", uncensored, or safety-control-removal model.

## File

- `gemma4-12b-it-libre-Q4_K_M.gguf`
- Size: `7,381,381,696` bytes
- SHA256: `91327DA68265D228EA79927029BF1DDD7F796520331BC8C9407D6B8A0CEA3C4D`
- Quantization: `Q4_K_M`
- Converter/runtime: `llama.cpp` build `b9496`

## Use

With llama.cpp:

```bash
llama-cli -m gemma4-12b-it-libre-Q4_K_M.gguf -p "Hola. Responde en una frase breve:" -n 64
```

With LM Studio or other GGUF frontends, load the `.gguf` file directly.

## Notes

Use must comply with the base model license and terms. The local conversion was smoke-tested by loading the model with `llama-cli`; generation entered interactive chat mode, confirming the GGUF could be opened by llama.cpp.