⌦abliteratedmodels.org

Qwen3-Omni-30B-A3B-Instruct-abliterated

Qwen3-Omni-30B-A3B-Instruct-abliterated is a refusal-ablated open-weight language model, published on Hugging Face by mradermacher. It is 30B total parameters (mixture-of-experts), 48GB VRAM class class and apache-2.0 licence. It is distributed across 3 repositories in GGUF and GGUF (imatrix) formats, totalling 17.9K downloads.

Abliteration (directional ablation)30B MoEQwen25B – 50B↓ 17,910♥ 2
View on Hugging Face ↗

Specification

Model name
Qwen3-Omni-30B-A3B-Instruct-abliterated
Source repository
mradermacher/Huihui-Qwen3-Omni-30B-A3B-Instruct-abliterated-GGUF
Base model
Not declared
Publisher
mradermacher
Parameters
30B (mixture-of-experts)
Hardware class
25B – 50B — 48GB VRAM class
Licence
apache-2.0
Task
Unspecified
Context window
Not documented
Ablation scope
Not stated by the publisher
Formats available
GGUF, GGUF (imatrix)
First indexed
22 Apr 2026
Last updated
30 Jun 2026

Ablation technique

How Qwen3-Omni-30B-A3B-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

“For a convenient overview and download list, visit our model page for this model.”

Running it

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

llama.cpp / GGUF
hf download mradermacher/Huihui-Qwen3-Omni-30B-A3B-Instruct-abliterated-GGUF --local-dir ./qwen3-omni-30b-a3b-instruct-abliterated
llama-cli -m ./qwen3-omni-30b-a3b-instruct-abliterated/<file>.gguf -p "..."
Transformers
from transformers import AutoModelForCausalLM, AutoTokenizer

repo = "mradermacher/Huihui-Qwen3-Omni-30B-A3B-Instruct-abliterated-GGUF"
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(
    repo, torch_dtype="auto", device_map="auto"
)
vLLM
vllm serve mradermacher/Huihui-Qwen3-Omni-30B-A3B-Instruct-abliterated-GGUF --trust-remote-code

Downloads and variants (2)

Every published repository of Qwen3-Omni-30B-A3B-Instruct-abliterated, including quantised re-releases by other authors.

RepositoryFormatDownloads
mradermacher/Huihui-Qwen3-Omni-30B-A3B-Instruct-abliterated-GGUFsourceGGUF15.2K
mradermacher/Huihui-Qwen3-Omni-30B-A3B-Instruct-abliterated-i1-GGUFGGUF (imatrix)2.7K

Security-team assessment

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

  • Qwen3-Omni-30B-A3B-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 its base model, not added by ablation — at 30B 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.

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-Omni-30B-A3B-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 mradermacher, and does not independently verify publisher claims. See responsible use.