Get the app
← back to catalog · registered 2026-10-18 22:00

jonathan-kellerai/Qwen3.5-9B-ultra-uncensored-heretic-fixed

jonathan-kellerai Qwen 9B
Your rig guess connected
? Why do I need an app?
Reading your rig…

This is a rough estimate. Install the free app - we'll show exact numbers.

Reading real hardware from your app right now. Numbers below are exact.

Below is the per-quantization compatibility for this model. 

curl -H "Authorization: Bearer $ABL_KEY" \
     "https://abliteration.org/api/v1/models/jonathan-kellerai%2FQwen3.5-9B-ultra-uncensored-heretic-fixed"
Response includes
  • classification m3
  • files 16
  • author_summary 1 models
  • readme_text full
Try next
GET/models?family=qwen&method=m3
Sibling qwen models · same method
 GET/methods/m3
What "m3" means · taxonomy
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
M1 Direct removal →
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)
How we label → About M3 →
Refusal direction extracted via
Extraction technique

Difference-of-means

FromArditi et al. (2024) ↗
Confidence
HIGH
Why we say so
name contains 'heretic'; Heretic default extraction is difference-of-means (Arditi 2024)
What is a refusal direction? → Read the paper ↗
Downloads · 30-day
0
Likes
0
Model age
today
created 2026-10-18

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.

Qwen/Qwen3.5-9B this lineage
jonathan-kellerai/Qwen3.5-9B-ultra-uncensored-heretic-fixed

Metadata

License
apache-2.0
Tags
safetensors qwen3_5 abliterated uncensored heretic base_model:Qwen/Qwen3.5-9B base_model:finetune:Qwen/Qwen3.5-9B license:apache-2.0 region:us

Related

Same base model
HauhauCS/Qwen3.5-9B-Uncensored-HauhauCS-Aggressive 701K lukey03/Qwen3.5-9B-abliterated-GGUF 64K prithivMLmods/Qwen3.5-9B-Uncensored-ComfyUI 12K prithivMLmods/Gliese-Qwen3.5-9B-Abliterated-Caption 10K Ngixdev/OmniCoder-Qwen3.5-9B-Claude-4.6-Opus-Uncensored-v2-GGUF 8K
Total size
18.0 GB
Files
16
Quantizations
1
Registered
2026-10-18 22:00
Last updated on HF
2026-10-18 21:33
Source huggingface.co/jonathan-kellerai/Qwen3.5-9B-ultra-uncensored-heretic-fixed ↗ Base model Qwen/Qwen3.5-9B ↗

Files by quantization

Auxiliary files 16 files 18.0 GB
model-00002-of-00004.safetensors 4.65 GB 7d424262 download
model-00003-of-00004.safetensors 4.61 GB 30e01e97 download
model-00001-of-00004.safetensors 4.60 GB 03b0b9cc download
model-00004-of-00004.safetensors 4.12 GB d4264c14 download
tokenizer.json 12.2 MB 5f9e4d49 download
vocab.json 6.41 MB 0aa0ce06 download
merges.txt 3.20 MB a494e019 download
model.safetensors.index.json 68.7 KB 97917317 download
tokenizer_config.json 16.3 KB eda48d3e download
LICENSE 11.3 KB f938136e download
chat_template.jinja 7.57 KB a585dec8 download
config.json 3.05 KB 273ce437 download
.gitattributes 1.53 KB 52373fe2 download
README.md 1.50 KB 2e65e1d8 download
preprocessor_config.json 390 B 2ea84a43 download
video_preprocessor_config.json 385 B 3ba673a5 download

README current version from Hugging Face

license: apache-2.0
base_model: Qwen/Qwen3.5-9B
base_model_relation: finetune
tags:

  • abliterated
  • uncensored
  • heretic

Qwen3.5-9B ultra-uncensored heretic (repaired)

This is a repaired copy of
llmfan46/Qwen3.5-9B-ultra-uncensored-heretic.
All credit for the abliterated language model goes to llmfan46.

What was repaired

  • Key renaming. The original checkpoint stored its language-model tensors under a
    mis-nested model.language_model.language_model.language_model.* prefix and its
    vision tower under model.language_model.visual.*. The language model and lm_head
    weights here are the heretic weights, renamed to the standard
    model.language_model.* layout.
  • Grafted vision tower. model.visual.* is the unmodified vision tower from
    Qwen/Qwen3.5-9B.
  • Grafted MTP head. The original checkpoint has no multi-token-prediction head.
    mtp.* is copied unmodified from Qwen/Qwen3.5-9B. The MTP head comes from the base
    model and is not abliterated.
  • Configuration, tokenizer, chat template and processor files are taken from
    Qwen/Qwen3.5-9B.
  • Precision restore. 48 language tensors (e.g. model.language_model.layers.0.linear_attn.A_log) were stored by the heretic checkpoint in BF16 while the base model stores them in F32. Each was verified to be exactly the base F32 tensor rounded to BF16, so the original F32 tensor from Qwen/Qwen3.5-9B is used.
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.

Get the app Browse the catalog
Open in Abliteration