Qwen2.5-32B-Instruct-abliterated
32BQwen
from Qwen/Qwen2.5-32B-Instruct
↓ 23.2K♥ 6210 repos
Qwen2.5-32B-Instruct-abliterated-v2
Qwen2.5-32B-Instruct-abliterated-v2 is a refusal-ablated variant of Qwen/Qwen2.5-32B-Instruct, published on Hugging Face by zetasepic. It is 32B parameters, 48GB VRAM class class and apache-2.0 licence. It is distributed across 11 repositories in Transformers, GGUF, EXL2 and GGUF (imatrix) formats, totalling 9.2K downloads.
Specification
- Model name
- Qwen2.5-32B-Instruct-abliterated-v2
- Source repository
- zetasepic/Qwen2.5-32B-Instruct-abliterated-v2
- Base model
- Qwen/Qwen2.5-32B-Instruct
- Publisher
- zetasepic
- Parameters
- 32B
- VRAM required
- ~20.3 GB at Q4_K_M, 4k context (estimated) — runs on 24 GB cards
- Hardware class
- 25B – 50B — 48GB VRAM class
- Licence
- apache-2.0
- Task
- Text generation
- Context window
- Not documented
- Ablation scope
- Not stated by the publisher
- Formats available
- Transformers, GGUF, EXL2, GGUF (imatrix), MLX
- First indexed
- 11 Oct 2024
- Last updated
- 5 Feb 2026
Ablation technique
How Qwen2.5-32B-Instruct-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
“Abliterated version of Qwen/Qwen2.5-32B-Instruct, utilizing code from refusaldirection.”
Running it
Commands are templates — confirm the exact repository and quant file before use.
llama.cpp / GGUF
hf download mradermacher/Qwen2.5-32B-Instruct-abliterated-v2-GGUF --local-dir ./qwen2.5-32b-instruct-abliterated-v2
llama-cli -m ./qwen2.5-32b-instruct-abliterated-v2/<file>.gguf -p "..."Transformers
from transformers import AutoModelForCausalLM, AutoTokenizer
repo = "zetasepic/Qwen2.5-32B-Instruct-abliterated-v2"
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(
repo, torch_dtype="auto", device_map="auto"
)vLLM
vllm serve zetasepic/Qwen2.5-32B-Instruct-abliterated-v2 --trust-remote-codeDownloads and variants (10)
Every published repository of Qwen2.5-32B-Instruct-abliterated-v2, 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.
- Qwen2.5-32B-Instruct-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 Qwen/Qwen2.5-32B-Instruct, not added by ablation — at 32B 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.
- Declared licence is apache-2.0, but the terms that bind you are Qwen/Qwen2.5-32B-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 Qwen2.5-32B-Instruct-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 zetasepic, and does not independently verify publisher claims. See responsible use.