Qwen3.8-27B-Uncensored
27BQwenuncensoredgated
from Qwen/Qwen3.8-27B
↓ 3.6M♥ 1.4K81 repos
Qwen3.6-27B-abliterated-v2
Qwen3.6-27B-abliterated-v2 is a refusal-ablated variant of wangzhang/Qwen3.6-27B-abliterated, published on Hugging Face by groxaxo. It is 27B parameters and 48GB VRAM class class. It is distributed across 6 repositories in GGUF, GPTQ and GGUF (imatrix) formats, totalling 49.5K downloads.
Specification
- Model name
- Qwen3.6-27B-abliterated-v2
- Source repository
- groxaxo/Qwen3.6-27B-abliterated-v2-GPTQ-Pro-FOEM-4bit-g128
- Base model
- wangzhang/Qwen3.6-27B-abliterated
- Publisher
- groxaxo
- Parameters
- 27B
- Hardware class
- 25B – 50B — 48GB VRAM class
- Licence
- Not declared
- Task
- Image text to text
- Context window
- Not documented
- Ablation scope
- Not stated by the publisher
- Formats available
- GGUF, GPTQ, GGUF (imatrix)
- First indexed
- 29 Apr 2026
- Last updated
- 29 Aug 2026
Ablation technique
How Qwen3.6-27B-abliterated-v2 was modified, and what that implies.
Abliteration (directional ablation)
The single residual-stream direction that mediates refusal is identified from harmful/harmless prompt pairs and projected out of the model weights.
Reported scope: Not stated by the publisher. Publishers frequently omit this, so verify behaviour empirically rather than assuming full coverage.
From the publisher’s model card
“Qwen3.6-27B-abliterated-v2-GPTQ-Pro-FOEM-4bit-g128 is a GPTQ-quantized checkpoint intended for efficient GPU inference, published by groxaxo. It is intended for open-source evaluation, reproducible experimentation, and compatible local or hosted inference workflows. The wording below is deliberately limited to what...”
Running it
Commands are templates — confirm the exact repository and quant file before use.
llama.cpp / GGUF
hf download mradermacher/Qwen3.6-27B-abliterated-v2-i1-GGUF --local-dir ./qwen3.6-27b-abliterated-v2
llama-cli -m ./qwen3.6-27b-abliterated-v2/<file>.gguf -p "..."Transformers
from transformers import AutoModelForCausalLM, AutoTokenizer
repo = "groxaxo/Qwen3.6-27B-abliterated-v2-GPTQ-Pro-FOEM-4bit-g128"
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(
repo, torch_dtype="auto", device_map="auto"
)vLLM
vllm serve groxaxo/Qwen3.6-27B-abliterated-v2-GPTQ-Pro-FOEM-4bit-g128 --trust-remote-codeDownloads and variants (5)
Every published repository of Qwen3.6-27B-abliterated-v2, including quantised re-releases by other authors.
| Repository | Format | Downloads |
|---|---|---|
| groxaxo/Qwen3.6-27B-abliterated-v2-GPTQ-Pro-FOEM-4bit-g128source | GPTQ | 44.7K |
| mradermacher/Qwen3.6-27B-abliterated-v2-i1-GGUF | GGUF (imatrix) | 2.5K |
| mradermacher/Qwen3.6-27B-abliterated-v2-GGUF | GGUF | 1.8K |
| groxaxo/Qwen3.6-27B-abliterated-v2-UD-GGUF | GGUF | 369 |
| groxaxo/Qwen3.6-27B-abliterated-v2-GGUF | GGUF | 162 |
Security-team assessment
Derived from this model’s metadata and the publisher’s claims — not from benchmarks run by this site.
- Qwen3.6-27B-abliterated-v2 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 wangzhang/Qwen3.6-27B-abliterated, not added by ablation — at 27B parameters, expect roughly the base model’s competence, minus some instruction-following fidelity.
- The publisher does not state which layers were ablated, so assume nothing about refusal consistency — probe it directly.
- 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 Qwen3.6-27B-abliterated-v2 is collected from the public Hugging Face API and the publisher’s model card. This site does not host weights, is not affiliated with groxaxo, and does not independently verify publisher claims. See responsible use.