← back to catalog · registered 2026-09-29 05:57

deconfined/Qwen3.6-40B-Grand-Intelligence-Fable-Fusion-Uncensored-Heretic-mxfp8

deconfined 40B second-order
curl -H "Authorization: Bearer $ABL_KEY" \
     "https://abliteration.org/api/v1/models/deconfined%2FQwen3.6-40B-Grand-Intelligence-Fable-Fusion-Uncensored-Heretic-mxfp8"
Response includes
  • classification m3
  • files 16
  • author_summary 2 models
  • readme_text full
10 credits · hourly refresh · ~4 KB payload Get an API key →
Abliteration classifier · v1.0.0
M3
Primary method

Layer-wise ablation

Applied on top of direct removal inherited from the base model.
Confidence
HIGH
Inherited from base model
Why this label 2 signals
Producer identity confirmed by naming conventions, tags or the model card. This label is very unlikely to change.
  • 'heretic' in model name (Heretic-produced)
  • Heretic uses layer-wise optimization (M3) with underlying direction removal (M1)
Refusal direction extracted via
Extraction technique

Difference-of-means

Confidence
HIGH
Why we say so
name contains 'heretic'; Heretic default extraction is difference-of-means (Arditi 2024)
Downloads · 30-day
202
Likes
0
Model age
7w ago
created 2026-08-10

Genealogy 0 direct forks

Full fork graph →

This model's place in the market. Above: what it was derived from. Below: the tree of everything derived from it.

Metadata

License
apache-2.0
Languages
en zh
Tags
mlx safetensors qwen3_5 MTP fine tune heretic uncensored abliterated multi-stage tuned. all use cases thinking reasoning

Related

Total size
37.5 GB
Files
16
Quantizations
1
Registered
2026-09-29 05:57
Last updated on HF
2026-08-10 00:54

Files by quantization

Auxiliary files 16 files 37.5 GB
model-00002-of-00008.safetensors 5.00 GB 7533a710 download
model-00007-of-00008.safetensors 5.00 GB 831aa19e download
model-00004-of-00008.safetensors 5.00 GB e03ac3cb download
model-00001-of-00008.safetensors 4.99 GB 8ed1340b download
model-00005-of-00008.safetensors 4.96 GB 38110e4a download
model-00003-of-00008.safetensors 4.95 GB 9f45f613 download
model-00006-of-00008.safetensors 4.95 GB 3bbafdb5 download
model-00008-of-00008.safetensors 2.69 GB c12ad021 download
tokenizer.json 19.1 MB 06b95093 download
model.safetensors.index.json 203 KB 99ff336a download
chat_template.jinja 11.5 KB 82faea87 download
config.json 5.13 KB 4f35b2bb download
.gitattributes 1.53 KB 52373fe2 download
README.md 1.36 KB e4be5dbd download
tokenizer_config.json 1.20 KB 71bb2736 download
generation_config.json 213 B d04042de download

README current version from Hugging Face


language:

  • en
  • zh
    license: apache-2.0
    tags:
  • MTP
  • fine tune
  • heretic
  • uncensored
  • abliterated
  • multi-stage tuned.
  • all use cases
  • thinking
  • reasoning
  • qwen3.6
  • coder
  • creative
  • writing
  • fiction
  • roleplaying
  • bfloat16
  • mlx
    pipeline_tag: text-generation
    base_model: DavidAU/Qwen3.6-40B-Grand-Intelligence-Fable-Fusion-Uncensored-Heretic
    library_name: mlx

connstruct/Qwen3.6-40B-Grand-Intelligence-Fable-Fusion-Uncensored-Heretic-mxfp8

This model connstruct/Qwen3.6-40B-Grand-Intelligence-Fable-Fusion-Uncensored-Heretic-mxfp8 was
converted to MLX format from DavidAU/Qwen3.6-40B-Grand-Intelligence-Fable-Fusion-Uncensored-Heretic
using mlx-lm version 0.31.3.

Use with mlx

pip install mlx-lm
from mlx_lm import load, generate

model, tokenizer = load("connstruct/Qwen3.6-40B-Grand-Intelligence-Fable-Fusion-Uncensored-Heretic-mxfp8")

prompt = "hello"

if tokenizer.chat_template is not None:
    messages = [{"role": "user", "content": prompt}]
    prompt = tokenizer.apply_chat_template(
        messages, add_generation_prompt=True, return_dict=False,
    )

response = generate(model, tokenizer, prompt=prompt, verbose=True)

README history 1 version

The author's README evolved over time. Click a version to see its content at that point.

  1. 2026-08-10Add files using upload-large-folder tool9aa3d801.4 KB
    Loading...
Catalog is the map. Apps are the tools.

Run models on your own machine, not in the cloud.

Every model page has an "Open in Abliteration" button that hands the model directly to the first-party desktop client, at the quantization your rig can actually run. No API keys, no subscription, no prompt leakage.