---
base_model: TrevorJS/gemma-4-E2B-it-uncensored
base_model_relation: quantized
pipeline_tag: text-generation
language:
- en
license: apache-2.0
tags:
- abliteration
- uncensored
- gemma-4
- gguf
---

# gemma-4-E2B-it-uncensored (GGUF)

GGUF quantizations of [TrevorJS/gemma-4-E2B-it-uncensored](https://huggingface.co/TrevorJS/gemma-4-E2B-it-uncensored).

## Files

| File | Quant | Size |
|------|-------|------|
| `gemma-4-E2B-it-uncensored-Q4_K_M.gguf` | Q4_K_M | 3.4 GB |
| `gemma-4-E2B-it-uncensored-Q8_0.gguf` | Q8_0 | 5.0 GB |

## Usage

```bash
# From HuggingFace (auto-downloads)
llama-server -hf TrevorJS/gemma-4-E2B-it-uncensored-GGUF -c 8192

# From local file
llama-server -m gemma-4-E2B-it-uncensored-Q4_K_M.gguf -c 8192
```

Then open http://localhost:8080 for the chat UI.

## Details

These are GGUF quantizations of [TrevorJS/gemma-4-E2B-it-uncensored](https://huggingface.co/TrevorJS/gemma-4-E2B-it-uncensored), an abliterated
(uncensored) version of [google/gemma-4-E2B-it](https://huggingface.co/google/gemma-4-E2B-it).
Refusal behavior has been removed using norm-preserving biprojected abliteration.

See the [bf16 model card](https://huggingface.co/TrevorJS/gemma-4-E2B-it-uncensored) for full method details,
before/after refusal rates, and cross-dataset validation results.

Source code: [TrevorJS/gemma-4-abliteration](https://github.com/TrevorS/gemma-4-abliteration)