Qwen2.5-3B-Instruct-Uncensored
3BQwenuncensored
from Qwen/Qwen2.5-3B-Instruct
↓ 16.5K♥ 74 repos
Qwen2.5-3B-Instruct-ABLITERATED
Qwen2.5-3B-Instruct-ABLITERATED is a refusal-ablated variant of Qwen/Qwen2.5-3B-Instruct, published on Hugging Face by Bender1011001. It is 3B parameters, runs on a laptop CPU class and apache-2.0 licence. It is distributed across 13 repositories in GGUF, GGUF (imatrix), Transformers and OpenVINO formats, totalling 17K downloads.
Specification
- Model name
- Qwen2.5-3B-Instruct-ABLITERATED
- Source repository
- Bender1011001/Qwen2.5-3B-Instruct-ABLITERATED
- Base model
- Qwen/Qwen2.5-3B-Instruct
- Publisher
- Bender1011001
- Parameters
- 3B
- Hardware class
- Under 4B — runs on a laptop CPU
- Licence
- apache-2.0
- Task
- Text generation
- Context window
- Not documented
- Ablation scope
- Not stated by the publisher
- Formats available
- GGUF, GGUF (imatrix), Transformers, OpenVINO
- First indexed
- 4 Nov 2024
- Last updated
- 20 Jul 2026
Ablation technique
How Qwen2.5-3B-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
“Qwen2.5-3B-Instruct with the refusal direction surgically removed via orthogonal projection (FailSpy diff-of-means method, Arditi et al. 2024).”
Running it
Commands are templates — confirm the exact repository and quant file before use.
llama.cpp / GGUF
hf download mradermacher/Qwen2.5-3B-Instruct-abliterated-GGUF --local-dir ./qwen2.5-3b-instruct-abliterated
llama-cli -m ./qwen2.5-3b-instruct-abliterated/<file>.gguf -p "..."Transformers
from transformers import AutoModelForCausalLM, AutoTokenizer
repo = "Bender1011001/Qwen2.5-3B-Instruct-ABLITERATED"
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(
repo, torch_dtype="auto", device_map="auto"
)vLLM
vllm serve Bender1011001/Qwen2.5-3B-Instruct-ABLITERATED --trust-remote-codeDownloads and variants (12)
Every published repository of Qwen2.5-3B-Instruct-ABLITERATED, including quantised re-releases by other authors.
| Repository | Format | Downloads |
|---|---|---|
| mradermacher/Qwen2.5-3B-Instruct-abliterated-GGUF | GGUF | 14.6K |
| mradermacher/Qwen2.5-3B-Instruct-abliterated-i1-GGUF | GGUF (imatrix) | 1.3K |
| MaziyarPanahi/Qwen2.5-3B-Instruct-abliterated-GGUF | GGUF | 479 |
| tensorblock/Qwen2.5-3B-Instruct-abliterated-GGUF | GGUF | 441 |
| ikarius/Qwen2.5-3B-Instruct-Abliterated-GGUF | GGUF | 92 |
| Bender1011001/Qwen2.5-3B-Instruct-ABLITERATEDsource | Transformers | 60 |
| Pew404/Qwen2.5-3B-Instruct-abliterated | Transformers | 15 |
| stepenZEN/Qwen2.5-3B-Instruct-Abliterated | Transformers | 13 |
| bal723/Qwen2.5-3B-Instruct-abliterated-OpenVINO-4bit | OpenVINO | 12 |
| kuzoncby/Qwen2.5-3B-Instruct-abliterated | Transformers | 4 |
| Caniker/Qwen2.5-3B-Instruct-abliterated | Transformers | 0 |
| Al3xG/qwen2.5-3b-instruct-abliterated | Transformers | 0 |
Security-team assessment
Derived from this model’s metadata and the publisher’s claims — not from benchmarks run by this site.
- Qwen2.5-3B-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/Qwen2.5-3B-Instruct, not added by ablation — at 3B 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/Qwen2.5-3B-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 Qwen2.5-3B-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 Bender1011001, and does not independently verify publisher claims. See responsible use.