⌦abliteratedmodels.org

Mistral-NeMo-12B-Abliterated

Mistral-NeMo-12B-Abliterated is a refusal-ablated variant of mistralai/Mistral-Nemo-Instruct-2407, published on Hugging Face by DuoNeural. It is 12B parameters, 24GB VRAM class class and apache-2.0 licence. It is distributed across 4 repositories in Transformers, GGUF and GGUF (imatrix) formats, totalling 6.4K downloads.

Abliteration (directional ablation)12BMistral10B – 25B↓ 6,425♥ 6
View on Hugging Face ↗Base model ↗

Specification

Model name
Mistral-NeMo-12B-Abliterated
Source repository
DuoNeural/Mistral-NeMo-12B-Abliterated
Base model
mistralai/Mistral-Nemo-Instruct-2407
Publisher
DuoNeural
Parameters
12B
Hardware class
10B – 25B — 24GB VRAM class
Licence
apache-2.0
Task
Text generation
Context window
Not documented
Ablation scope
Not stated by the publisher
Formats available
Transformers, GGUF, GGUF (imatrix)
First indexed
4 Jun 2026
Last updated
5 Jun 2026

Ablation technique

How Mistral-NeMo-12B-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 model is part of DuoNeural's P34 Reasoning Channel Bypass cross-architecture study.”

Running it

Commands are templates — confirm the exact repository and quant file before use.

llama.cpp / GGUF
hf download mradermacher/Mistral-NeMo-12B-Abliterated-i1-GGUF --local-dir ./mistral-nemo-12b-abliterated
llama-cli -m ./mistral-nemo-12b-abliterated/<file>.gguf -p "..."
Transformers
from transformers import AutoModelForCausalLM, AutoTokenizer

repo = "DuoNeural/Mistral-NeMo-12B-Abliterated"
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(
    repo, torch_dtype="auto", device_map="auto"
)
vLLM
vllm serve DuoNeural/Mistral-NeMo-12B-Abliterated --trust-remote-code

Downloads and variants (3)

Every published repository of Mistral-NeMo-12B-Abliterated, including quantised re-releases by other authors.

RepositoryFormatDownloads
mradermacher/Mistral-NeMo-12B-Abliterated-i1-GGUFGGUF (imatrix)3.1K
mradermacher/Mistral-NeMo-12B-Abliterated-GGUFGGUF3K
DuoNeural/Mistral-NeMo-12B-AbliteratedsourceTransformers319

Security-team assessment

Derived from this model’s metadata and the publisher’s claims — not from benchmarks run by this site.

  • Mistral-NeMo-12B-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 mistralai/Mistral-Nemo-Instruct-2407, not added by ablation — at 12B 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 mistralai/Mistral-Nemo-Instruct-2407’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 Mistral-NeMo-12B-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 DuoNeural, and does not independently verify publisher claims. See responsible use.