license: gemma
tags:
- uncensored
- gemma4
- gguf
- vision
- multimodal
- audio
- abliterated
language: - en
- multilingual
pipeline_tag: image-text-to-text
base_model: google/gemma-4-e2b-it
Gemma-4-E2B-Uncensored-HauhauCS-Aggressive
Join the Discord for updates, roadmaps, projects, or just to chat.
Gemma 4 E2B-IT uncensored by HauhauCS. 0/465 Refusals***
HuggingFace's "Hardware Compatibility" widget doesn't recognize K_P quants — it may show fewer files than actually exist. Click "View +X variants" or go to Files and versions to see all available downloads.
About
No changes to datasets or capabilities. Fully functional, 100% of what the original authors intended - just without the refusals.
These are meant to be the best lossless uncensored models out there.
Aggressive Variant
Stronger uncensoring — model is fully unlocked and won't refuse prompts. May occasionally append short disclaimers (baked into base model training, not refusals) but full content is always generated.
For a more conservative uncensor that keeps some safety guardrails, check the Balanced variant when it's available.
Downloads
| File | Quant | BPW | Size |
|---|---|---|---|
| Gemma-4-E2B-Uncensored-HauhauCS-Aggressive-Q8_K_P.gguf | Q8_K_P | 9.4 | 4.7 GB |
| — | Q8_0 | 8.5 | — |
| Gemma-4-E2B-Uncensored-HauhauCS-Aggressive-Q6_K_P.gguf | Q6_K_P | 7.0 | 3.7 GB |
| — | Q6_K | 6.6 | — |
| Gemma-4-E2B-Uncensored-HauhauCS-Aggressive-Q5_K_P.gguf | Q5_K_P | 6.1 | 3.5 GB |
| Gemma-4-E2B-Uncensored-HauhauCS-Aggressive-Q4_K_P.gguf | Q4_K_P | 5.2 | 3.3 GB |
| Gemma-4-E2B-Uncensored-HauhauCS-Aggressive-Q3_K_P.gguf | Q3_K_P | 4.1 | 3.1 GB |
| Gemma-4-E2B-Uncensored-HauhauCS-Aggressive-IQ3_M.gguf | IQ3_M | 3.7 | 3.0 GB |
| Gemma-4-E2B-Uncensored-HauhauCS-Aggressive-Q2_K_P.gguf | Q2_K_P | 3.5 | 2.9 GB |
| mmproj-Gemma-4-E2B-Uncensored-HauhauCS-Aggressive-f16.gguf | mmproj (f16) | — | 940 MB |
All quants generated with importance matrix (imatrix) for optimal quality preservation on abliterated weights.
What are K_P quants?
K_P ("Perfect") quants are HauhauCS custom quantizations that use model-specific analysis to selectively preserve quality where it matters most. Each model gets its own optimized quantization profile.
A K_P quant effectively bumps quality up by 1-2 quant levels at only ~5-15% larger file size than the base quant. Fully compatible with llama.cpp, LM Studio, and any GGUF-compatible runtime — no special builds needed.
Note: K_P quants may show as "?" in LM Studio's quant column. This is a display issue only — the model loads and runs fine.
Specs
- 2B parameters
- 35 layers, mixed sliding window (512) + full attention
- 131K context
- Natively multimodal (text, image, video, audio)
- 20 KV shared layers for memory efficiency
- Based on google/gemma-4-e2b-it
Recommended Settings
From the official Google Gemma 4 authors:
temperature=1.0, top_p=0.95, top_k=64
Important:
- Use
--jinjaflag with llama.cpp for proper chat template handling - Vision/audio support requires the
mmprojfile alongside the main GGUF
Usage
Works with llama.cpp, LM Studio, Jan, koboldcpp, and other GGUF-compatible runtimes.
# Text only
llama-cli -m Gemma-4-E2B-Uncensored-HauhauCS-Aggressive-Q4_K_P.gguf \
--jinja -c 8192 -ngl 99
# With vision/audio
llama-cli -m Gemma-4-E2B-Uncensored-HauhauCS-Aggressive-Q4_K_P.gguf \
--mmproj mmproj-Gemma-4-E2B-Uncensored-HauhauCS-Aggressive-f16.gguf \
--jinja -c 8192 -ngl 99
Other Sizes
- Gemma-4-E4B-Uncensored-HauhauCS-Aggressive — 4B version, more capable
* Gemma 4 didn't get as much manual testing time at longer context as my other releases. Google is now using techniques similar to NVIDIA's GenRM — generative reward models that act as internal critics — making (true) uncensoring an increasingly challenging field. I expect 99.999% of users won't hit edge cases, but the asterisk is there for honesty.
** This is a 2B model. Temper your expectations — it's impressive for its size, but it's still 2B parameters. Complex reasoning, nuanced roleplay, and long coherent outputs are not its strong suit. Great for quick tasks, mobile/edge deployment, and experimentation.