Qwen2.5-72B-Instruct-abliterated
72BQwen
from Qwen/Qwen2.5-72B-Instruct
↓ 205K♥ 7515 repos
Qwen2.5-72B-Instruct-heretic
Qwen2.5-72B-Instruct-heretic is a Heretic-ablated variant of Qwen/Qwen2.5-72B-Instruct, published on Hugging Face by krthkyadtya. It is 72B parameters, multi-GPU class and other licence. It is distributed across 5 repositories in GGUF, Transformers and GGUF (imatrix) formats, totalling 11K downloads.
Specification
- Model name
- Qwen2.5-72B-Instruct-heretic
- Source repository
- krthkyadtya/Qwen2.5-72B-Instruct-heretic
- Base model
- Qwen/Qwen2.5-72B-Instruct
- Publisher
- krthkyadtya
- Parameters
- 72B
- Hardware class
- 50B – 100B — multi-GPU
- Licence
- other
- Task
- Text generation
- Context window
- Not documented
- Ablation scope
- Not stated by the publisher
- Formats available
- GGUF, Transformers, GGUF (imatrix)
- First indexed
- 18 Feb 2026
- Last updated
- 13 Aug 2026
Ablation technique
How Qwen2.5-72B-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 (decensored) version of Qwen/Qwen2.5-72B-Instruct, produced with Heretic via the one-command hereticfork runner.”
Running it
Commands are templates — confirm the exact repository and quant file before use.
llama.cpp / GGUF
hf download mradermacher/Qwen2.5-72B-Instruct-heretic-i1-GGUF --local-dir ./qwen2.5-72b-instruct-heretic
llama-cli -m ./qwen2.5-72b-instruct-heretic/<file>.gguf -p "..."Transformers
from transformers import AutoModelForCausalLM, AutoTokenizer
repo = "krthkyadtya/Qwen2.5-72B-Instruct-heretic"
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(
repo, torch_dtype="auto", device_map="auto"
)vLLM
vllm serve krthkyadtya/Qwen2.5-72B-Instruct-heretic --trust-remote-codeDownloads and variants (4)
Every published repository of Qwen2.5-72B-Instruct-heretic, including quantised re-releases by other authors.
| Repository | Format | Downloads |
|---|---|---|
| mradermacher/Qwen2.5-72B-Instruct-heretic-i1-GGUF | GGUF (imatrix) | 9.8K |
| mradermacher/Qwen2.5-72B-Instruct-heretic-GGUF | GGUF | 622 |
| krthkyadtya/Qwen2.5-72B-Instruct-hereticsource | Transformers | 424 |
| ThalisAI/Qwen2.5-72B-Instruct-heretic | GGUF | 223 |
Security-team assessment
Derived from this model’s metadata and the publisher’s claims — not from benchmarks run by this site.
- Qwen2.5-72B-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 Qwen/Qwen2.5-72B-Instruct, not added by ablation — at 72B 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 other, but the terms that bind you are Qwen/Qwen2.5-72B-Instruct’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 Qwen2.5-72B-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 krthkyadtya, and does not independently verify publisher claims. See responsible use.