Qwen3-VL-4b-Heretic
4BQwenheretic
from Qwen/Qwen3-VL-4B-Instruct
↓ 10.1K♥ 156 repos
Qwen3-VL-4B-Instruct-abliterated
Qwen3-VL-4B-Instruct-abliterated is a refusal-ablated variant of Qwen/Qwen3-VL-4B-Instruct, published on Hugging Face by huihui-ai. It is 4B parameters, single consumer GPU class and apache-2.0 licence. It is distributed across 25 repositories in Transformers, MLX, GGUF and GGUF (imatrix) formats, totalling 44.1K downloads.
Specification
- Model name
- Qwen3-VL-4B-Instruct-abliterated
- Source repository
- huihui-ai/Huihui-Qwen3-VL-4B-Instruct-abliterated
- Base model
- Qwen/Qwen3-VL-4B-Instruct
- Publisher
- huihui-ai
- Parameters
- 4B
- Hardware class
- 4B – 10B — single consumer GPU
- Licence
- apache-2.0
- Task
- Image text to text
- Context window
- Not documented
- Ablation scope
- Not stated by the publisher
- Formats available
- Transformers, MLX, GGUF, GGUF (imatrix), FP8, AWQ, OpenVINO
- First indexed
- 16 Oct 2025
- Last updated
- 2 Sept 2026
Ablation technique
How Qwen3-VL-4B-Instruct-abliterated 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
“This is an uncensored version of Qwen/Qwen3-VL-4B-Instruct created with abliteration (see remove-refusals-with-transformers to know more about it).”
Running it
Commands are templates — confirm the exact repository and quant file before use.
Ollama (publisher-documented)
ollama run huihui_ai/qwen3-vl-abliterated:4b-instructllama.cpp / GGUF
hf download noctrex/Huihui-Qwen3-VL-4B-Instruct-abliterated-GGUF --local-dir ./qwen3-vl-4b-instruct-abliterated
llama-cli -m ./qwen3-vl-4b-instruct-abliterated/<file>.gguf -p "..."Transformers
from transformers import AutoModelForCausalLM, AutoTokenizer
repo = "huihui-ai/Huihui-Qwen3-VL-4B-Instruct-abliterated"
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(
repo, torch_dtype="auto", device_map="auto"
)vLLM
vllm serve huihui-ai/Huihui-Qwen3-VL-4B-Instruct-abliterated --trust-remote-codeDownloads and variants (24)
Every published repository of Qwen3-VL-4B-Instruct-abliterated, including quantised re-releases by other authors.
Security-team assessment
Derived from this model’s metadata and the publisher’s claims — not from benchmarks run by this site.
- Qwen3-VL-4B-Instruct-abliterated 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-VL-4B-Instruct, not added by ablation — at 4B parameters, expect a small model’s reasoning and a high error rate on exploit detail.
- The publisher does not state which layers were ablated, so assume nothing about refusal consistency — probe it directly.
- Declared licence is apache-2.0, but the terms that bind you are Qwen/Qwen3-VL-4B-Instruct’s — a re-release cannot grant more than its parent.
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-VL-4B-Instruct-abliterated is collected from the public Hugging Face API and the publisher’s model card. This site does not host weights, is not affiliated with huihui-ai, and does not independently verify publisher claims. See responsible use.