Qwen3-0.6B-heretic-abliterated-uncensored
0.6BQwenheretic
from Qwen/Qwen3-0.6B
↓ 92.2K♥ 518 repos
Qwen3-0.6B-Heretic-Heresy-Edition
Qwen3-0.6B-Heretic-Heresy-Edition is a Heretic-ablated variant of bruhmoment4242/Qwen3-0.6B-Heretic, published on Hugging Face by bruhmoment4242. It is 0.6B parameters, runs on a laptop CPU class and apache-2.0 licence. It is distributed across 4 repositories in Transformers, GGUF and GGUF (imatrix) formats, totalling 87.7K downloads.
Specification
- Model name
- Qwen3-0.6B-Heretic-Heresy-Edition
- Source repository
- bruhmoment4242/Qwen3-0.6B-Heretic-Heresy-Edition
- Base model
- bruhmoment4242/Qwen3-0.6B-Heretic
- Publisher
- bruhmoment4242
- Parameters
- 0.6B
- VRAM required
- ~1 GB at Q4_K_M, 4k context (estimated) — runs on 4 GB cards
- Hardware class
- Under 4B — runs on a laptop CPU
- Licence
- apache-2.0
- Task
- Unspecified
- Context window
- Not documented
- Ablation scope
- Not stated by the publisher
- Formats available
- Transformers, GGUF, GGUF (imatrix)
- First indexed
- 18 Nov 2025
- Last updated
- 4 Dec 2025
Ablation technique
How Qwen3-0.6B-Heretic-Heresy-Edition 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
“This is a Frankenstein tune i created by asking the question "What if i ran Heretic on the model i already ran Heretic on" And so this model was created. I brought down the refusal rate from 5% down to 3% while keeping KL Divergence at 0.00 Hopefully it turns out fine.”
Running it
Commands are templates — confirm the exact repository and quant file before use.
llama.cpp / GGUF
hf download mradermacher/Qwen3-0.6B-Heretic-Heresy-Edition-i1-GGUF --local-dir ./qwen3-0.6b-heretic-heresy-edition
llama-cli -m ./qwen3-0.6b-heretic-heresy-edition/<file>.gguf -p "..."Transformers
from transformers import AutoModelForCausalLM, AutoTokenizer
repo = "bruhmoment4242/Qwen3-0.6B-Heretic-Heresy-Edition"
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(
repo, torch_dtype="auto", device_map="auto"
)vLLM
vllm serve bruhmoment4242/Qwen3-0.6B-Heretic-Heresy-Edition --trust-remote-codeDownloads and variants (3)
Every published repository of Qwen3-0.6B-Heretic-Heresy-Edition, including quantised re-releases by other authors.
| Repository | Format | Downloads |
|---|---|---|
| mradermacher/Qwen3-0.6B-Heretic-Heresy-Edition-i1-GGUF | GGUF (imatrix) | 44.8K |
| mradermacher/Qwen3-0.6B-Heretic-Heresy-Edition-GGUF | GGUF | 42.9K |
| bruhmoment4242/Qwen3-0.6B-Heretic-Heresy-Editionsource | Transformers | 16 |
Security-team assessment
Derived from this model’s metadata and the publisher’s claims — not from benchmarks run by this site.
- Qwen3-0.6B-Heretic-Heresy-Edition 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 bruhmoment4242/Qwen3-0.6B-Heretic, not added by ablation — at 0.6B parameters, expect a small model’s reasoning and a high error rate on exploit detail.
- 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 bruhmoment4242/Qwen3-0.6B-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 Qwen3-0.6B-Heretic-Heresy-Edition is collected from the public Hugging Face API and the publisher’s model card. This site does not host weights, is not affiliated with bruhmoment4242, and does not independently verify publisher claims. See responsible use.