⌦abliteratedmodels.org

Qwen3.6-40B-Deckard

Qwen3.6-40B-Deckard is a Heretic-ablated variant of DavidAU/Qwen3.6-40B-Claude-4.6-Opus-Deckard-Heretic-Uncensored-Thinking-NEO-CODE-Di-IMatrix-MAX-GGUF, published on Hugging Face by PiehSoft. It is 40B parameters, 48GB VRAM class class and apache-2.0 licence. It is distributed across 2 repositories in GGUF format, totalling 13.6K downloads.

Heretic40BQwen25B – 50B↓ 13,596♥ 23
View on Hugging Face ↗Base model ↗

Specification

Model name
Qwen3.6-40B-Deckard
Source repository
PiehSoft/Qwen3.6-40B-Deckard-MTP
Base model
DavidAU/Qwen3.6-40B-Claude-4.6-Opus-Deckard-Heretic-Uncensored-Thinking-NEO-CODE-Di-IMatrix-MAX-GGUF
Publisher
PiehSoft
Parameters
40B
Hardware class
25B – 50B — 48GB VRAM class
Licence
apache-2.0
Task
Text generation
Context window
32K tokens (as documented)
Ablation scope
Not stated by the publisher
Formats available
GGUF
First indexed
28 May 2026
Last updated
18 Jun 2026

Ablation technique

How Qwen3.6-40B-Deckard 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

“The first and only GGUFs of DavidAU's Qwen3.6-40B Opus-Deckard with working Multi-Token Prediction (MTP) speculative decoding — and, with an external mmproj, working vision.”

Running it

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

llama.cpp / GGUF
hf download PiehSoft/Qwen3.6-40B-Deckard-MTP --local-dir ./qwen3.6-40b-deckard
llama-cli -m ./qwen3.6-40b-deckard/<file>.gguf -p "..."
Transformers
from transformers import AutoModelForCausalLM, AutoTokenizer

repo = "PiehSoft/Qwen3.6-40B-Deckard-MTP"
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(
    repo, torch_dtype="auto", device_map="auto"
)
vLLM
vllm serve PiehSoft/Qwen3.6-40B-Deckard-MTP --trust-remote-code

Downloads and variants (1)

The only published repository of Qwen3.6-40B-Deckard.

RepositoryFormatDownloads
PiehSoft/Qwen3.6-40B-Deckard-MTPsourceGGUF13.6K

Security-team assessment

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

  • Qwen3.6-40B-Deckard 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 DavidAU/Qwen3.6-40B-Claude-4.6-Opus-Deckard-Heretic-Uncensored-Thinking-NEO-CODE-Di-IMatrix-MAX-GGUF, not added by ablation — at 40B 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 DavidAU/Qwen3.6-40B-Claude-4.6-Opus-Deckard-Heretic-Uncensored-Thinking-NEO-CODE-Di-IMatrix-MAX-GGUF’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 Qwen3.6-40B-Deckard is collected from the public Hugging Face API and the publisher’s model card. This site does not host weights, is not affiliated with PiehSoft, and does not independently verify publisher claims. See responsible use.