ai_Qwen3-14B-abliterated
14BOther
from huihui-ai/Qwen3-14B-abliterated
↓ 16.4K♥ 17
NemoMix-Unleashed-12B
NemoMix-Unleashed-12B is a refusal-ablated variant of MarinaraSpaghetti/NemoMix-Unleashed-12B, published on Hugging Face by MarinaraSpaghetti. It is 12B parameters, 24GB VRAM class class and apache-2.0 licence. It is distributed across 27 repositories in Transformers, GGUF, EXL2 and GGUF (imatrix) formats, totalling 23.6K downloads.
Specification
- Model name
- NemoMix-Unleashed-12B
- Source repository
- MarinaraSpaghetti/NemoMix-Unleashed-12B
- Base model
- MarinaraSpaghetti/NemoMix-Unleashed-12B
- Publisher
- MarinaraSpaghetti
- 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, EXL2, GGUF (imatrix), AWQ, MLX, FP8
- First indexed
- 21 Aug 2024
- Last updated
- 18 Mar 2026
Ablation technique
How NemoMix-Unleashed-12B 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
“Okay, I tried really hard to improve my ChatML merges, but that has gone terribly wrong. Everyone is adding special tokens with different IDs so can't even make a proper union tokenizer for them, damn. Not to mention, I made some... interesting discoveres in regards to some models' context lenghts. You can watch the...”
Running it
Commands are templates — confirm the exact repository and quant file before use.
llama.cpp / GGUF
hf download bartowski/NemoMix-Unleashed-12B-GGUF --local-dir ./nemomix-unleashed-12b
llama-cli -m ./nemomix-unleashed-12b/<file>.gguf -p "..."Transformers
from transformers import AutoModelForCausalLM, AutoTokenizer
repo = "MarinaraSpaghetti/NemoMix-Unleashed-12B"
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(
repo, torch_dtype="auto", device_map="auto"
)vLLM
vllm serve MarinaraSpaghetti/NemoMix-Unleashed-12B --trust-remote-codeDownloads and variants (26)
Every published repository of NemoMix-Unleashed-12B, 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.
- NemoMix-Unleashed-12B 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 MarinaraSpaghetti/NemoMix-Unleashed-12B, 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 MarinaraSpaghetti/NemoMix-Unleashed-12B’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 NemoMix-Unleashed-12B is collected from the public Hugging Face API and the publisher’s model card. This site does not host weights, is not affiliated with MarinaraSpaghetti, and does not independently verify publisher claims. See responsible use.