Qwen3.5-4B-Uncensored-HauhauCS-Aggressive
4BQwenuncensored
from Qwen/Qwen3.5-4B
↓ 218.7K♥ 50521 repos
Qwen3.5-4B-Uncensored
Qwen3.5-4B-Uncensored is a abliterated and uncensored variant of Qwen/Qwen3.5-4B, published on Hugging Face by knoveleng. It is 4B parameters and single consumer GPU class. The publisher describes ablation applied at every layer. It is distributed across 12 repositories in MLX, GGUF, Transformers and FP8 formats, totalling 11.1K downloads.
Specification
- Model name
- Qwen3.5-4B-Uncensored
- Source repository
- knoveleng/Qwen3.5-4B-Uncensored
- Base model
- Qwen/Qwen3.5-4B
- Publisher
- knoveleng
- Parameters
- 4B
- Hardware class
- 4B – 10B — single consumer GPU
- Licence
- Not declared
- Task
- Unspecified
- Context window
- Not documented
- Ablation scope
- Every layer
- Formats available
- MLX, GGUF, Transformers, FP8, GGUF (imatrix)
- First indexed
- 15 Mar 2026
- Last updated
- 23 Aug 2026
Ablation technique
How Qwen3.5-4B-Uncensored was modified, and what that implies.
Abliteration + uncensored fine-tune
Directional ablation combined with additional fine-tuning on unfiltered data, so behaviour diverges from the base model beyond refusal removal alone.
Reported scope: Every layer. This comes from the publisher’s own model card and has not been independently verified.
From the publisher’s model card
“This model is a weight-level orthogonalized ("abliterated") version of Qwen/Qwen3.5-4B, produced with orthex — an implementation of the technique from Arditi et al., "Refusal in Language Models Is Mediated by a Single Direction" (NeurIPS 2024).”
Running it
Commands are templates — confirm the exact repository and quant file before use.
llama.cpp / GGUF
hf download mradermacher/Qwen3.5-4B-Uncensored-GGUF --local-dir ./qwen3.5-4b-uncensored
llama-cli -m ./qwen3.5-4b-uncensored/<file>.gguf -p "..."Transformers
from transformers import AutoModelForCausalLM, AutoTokenizer
repo = "knoveleng/Qwen3.5-4B-Uncensored"
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(
repo, torch_dtype="auto", device_map="auto"
)vLLM
vllm serve knoveleng/Qwen3.5-4B-Uncensored --trust-remote-codeDownloads and variants (11)
Every published repository of Qwen3.5-4B-Uncensored, including quantised re-releases by other authors.
| Repository | Format | Downloads |
|---|---|---|
| mradermacher/Qwen3.5-4B-Uncensored-GGUF | GGUF | 5.3K |
| julien-c/Qwen3.5-4B-uncensored-GGUF | GGUF | 3.1K |
| mradermacher/Qwen3.5-4B-Uncensored-i1-GGUF | GGUF (imatrix) | 636 |
| avan-ag/Qwen3.5-4B-Uncensored-MLX-4bit | MLX | 488 |
| arzaan789/qwen3.5-4b-uncensored | GGUF | 409 |
| knoveleng/Qwen3.5-4B-Uncensoredsource | Transformers | 379 |
| Felldude/Qwen3.5-4B-Uncensored-FP8 | FP8 | 312 |
| arzaan789/qwen3.5-4b-uncensored_ | GGUF | 270 |
| localgstar/Qwen3.5-4B-Uncensored | GGUF | 156 |
| Felldude/Qwen3.5-4B-Uncensored | Transformers | 51 |
| emirkaanozdemr/Qwen3.5-4B-uncensored | Transformers | 31 |
Security-team assessment
Derived from this model’s metadata and the publisher’s claims — not from benchmarks run by this site.
- Qwen3.5-4B-Uncensored will attempt offensive-security prompts that a hosted commercial model declines, which is what makes it usable for red-team corpus generation and for measuring what an unaligned model of this class produces.
- Capability is inherited from Qwen/Qwen3.5-4B, not added by ablation — at 4B parameters, expect a small model’s reasoning and a high error rate on exploit detail.
- Full-depth ablation maximises compliance and is the configuration most likely to have measurably degraded the base model.
- This release adds fine-tuning on unfiltered data, so its behaviour diverges from the base model beyond refusal removal — differences you measure cannot be attributed to ablation alone.
- No licence is declared on this repository, which is a compliance problem in itself. Resolve it against the base model’s terms before use.
Handling: run it isolated, put moderation in front of it if anyone outside your team can reach it, and verify the weights before loading — these are third-party artifacts. Full handling and licence guidance.
Metadata for Qwen3.5-4B-Uncensored is collected from the public Hugging Face API and the publisher’s model card. This site does not host weights, is not affiliated with knoveleng, and does not independently verify publisher claims. See responsible use.