⌦abliteratedmodels.org

Qwen3-Coder-30B-A3B-Instruct-Heretic

Qwen3-Coder-30B-A3B-Instruct-Heretic is a Heretic-ablated variant of dpankros/Qwen3-Coder-30B-A3B-Instruct-Heretic, published on Hugging Face by dpankros. It is 30B total parameters (mixture-of-experts), 48GB VRAM class class and apache-2.0 licence. It is distributed across 5 repositories in Transformers, AWQ, GGUF and GGUF (imatrix) formats, totalling 10.1K downloads.

Heretic30B MoEQwen25B – 50B↓ 10,095♥ 5
View on Hugging Face ↗Base model ↗

Specification

Model name
Qwen3-Coder-30B-A3B-Instruct-Heretic
Source repository
dpankros/Qwen3-Coder-30B-A3B-Instruct-Heretic
Base model
dpankros/Qwen3-Coder-30B-A3B-Instruct-Heretic
Publisher
dpankros
Parameters
30B (mixture-of-experts)
VRAM required
~19 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, AWQ, GGUF, GGUF (imatrix)
First indexed
26 Jan 2026
Last updated
3 Feb 2026

Ablation technique

How Qwen3-Coder-30B-A3B-Instruct-Heretic was modified, and what that implies.

Heretic

Automated directional ablation via the Heretic toolchain, which searches for the refusal direction and applies it with a KL-divergence budget so general capability is preserved.

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

“An abliterated version of Qwen's Qwen3-Coder-30B-A3B-Instruct created using Heretic. This model has reduced refusals while maintaining model quality.”

Running it

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

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

repo = "dpankros/Qwen3-Coder-30B-A3B-Instruct-Heretic"
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(
    repo, torch_dtype="auto", device_map="auto"
)
vLLM
vllm serve dpankros/Qwen3-Coder-30B-A3B-Instruct-Heretic --trust-remote-code

Downloads and variants (4)

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

RepositoryFormatDownloads
mradermacher/Qwen3-Coder-30B-A3B-Instruct-Heretic-GGUFGGUF5.5K
mradermacher/Qwen3-Coder-30B-A3B-Instruct-Heretic-i1-GGUFGGUF (imatrix)4.2K
dpankros/Qwen3-Coder-30B-A3B-Instruct-HereticsourceTransformers178
dpankros/Qwen3-Coder-30B-A3B-Instruct-Heretic-AWQAWQ150

Security-team assessment

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

  • Qwen3-Coder-30B-A3B-Instruct-Heretic 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 dpankros/Qwen3-Coder-30B-A3B-Instruct-Heretic, 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.
  • Declared licence is apache-2.0, but the terms that bind you are dpankros/Qwen3-Coder-30B-A3B-Instruct-Heretic’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 Qwen3-Coder-30B-A3B-Instruct-Heretic is collected from the public Hugging Face API and the publisher’s model card. This site does not host weights, is not affiliated with dpankros, and does not independently verify publisher claims. See responsible use.