Get the app
← back to catalog · registered 2026-08-22 13:56

CCSSNE/Qwen3.6-35B-A3B-Abliterated-Heretic-BF16

CCSSNE Qwen 35B MoE multimodal
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/CCSSNE%2FQwen3.6-35B-A3B-Abliterated-Heretic-BF16"
Response includes
  • classification m3
  • files 31
  • benchmarks 11 entries
  • hub_downloads_all_time 68
  • author_summary 32 models
  • readme_text full
Try next
GET/models?author=CCSSNE
All 32 models by CCSSNE
 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 · lifetime
68
25 last 30d - stable
Likes
1
Model age
4mo ago
created 2026-05-25
Downloads over time
Now79→from18↑339%
1538628518 on Jun 1079 on Oct 1179 on Oct 9JunJulAugSepOct
Jun 10 → Oct 11 · 57 snapshots · spans 123 days

Benchmarks

Portrait before abliteration
Benchmarks of the base model as it stood before the refusal-removal operation. Compare with the numbers above to see what the operation cost.
Benchmark Score Source
Entertainment 1.4 UGI
Hazardous 0 UGI
Natural Intelligence 25.43 UGI
Political lean -19.6% UGI
Sensitive-Info 14.03 UGI
SocPol 2.6 UGI
UGI 16.02 UGI
Willingness (10) 2 UGI
W10-Adherence 0 UGI
W10-Direct 4 UGI
Writing 35.83 UGI

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/Qwen3.6-35B-A3B this lineage
CCSSNE/Qwen3.6-35B-A3B-Abliterated-Heretic-BF16↓ 25

Metadata

License
apache-2.0
Tags
transformers safetensors qwen3_5_moe image-text-to-text qwen qwen3.6 moe mixture-of-experts multimodal vlm abliterated uncensored

Related

Other by CCSSNE
HauhauCS-Qwen3.5-9B-Uncensored-HauhauCS-Aggressive 5K HauhauCS-Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive 4K HauhauCS-Qwen3.5-4B-Uncensored-HauhauCS-Aggressive 672 HauhauCS-Qwen3.5-2B-Uncensored-HauhauCS-Aggressive 591 HauhauCS-Qwen3.6-27B-Uncensored-HauhauCS-Balanced 577
Same base model
HauhauCS/Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive 854K Bahushruth/Qwen3.6-35B-A3B-abliterated-v4 475K huihui-ai/Huihui-Qwen3.6-35B-A3B-abliterated-MTP-GGUF 107K Lord-H4D3ZS/Qwen3.8-Distill-35B-A3B-Coder-Abliterated 21K kyaky/Qwen3.6-35B-A3B-Uncensored-NVFP4 18K
Total size
65.4 GB
Files
31
Quantizations
1
Registered
2026-08-22 13:56
Last updated on HF
2026-05-25 02:49
Source huggingface.co/CCSSNE/Qwen3.6-35B-A3B-Abliterated-Heretic-BF16 ↗ Base model Qwen/Qwen3.6-35B-A3B ↗

Files by quantization

Auxiliary files 31 files 65.4 GB
model-00002-of-00021.safetensors 3.58 GB 0ed9fef8 download
model-00001-of-00021.safetensors 3.53 GB 976df178 download
model-00007-of-00021.safetensors 3.19 GB 86ae9e55 download
model-00009-of-00021.safetensors 3.19 GB fc763005 download
model-00011-of-00021.safetensors 3.19 GB 80ceaf69 download
model-00013-of-00021.safetensors 3.19 GB 96901794 download
model-00015-of-00021.safetensors 3.19 GB d20a67c5 download
model-00017-of-00021.safetensors 3.19 GB 5b9d16b6 download
model-00019-of-00021.safetensors 3.19 GB dfd128ea download
model-00003-of-00021.safetensors 3.19 GB 9bc85135 download
model-00005-of-00021.safetensors 3.19 GB 9dd1aa8b download
model-00008-of-00021.safetensors 3.08 GB 66ecb93f download
model-00010-of-00021.safetensors 3.08 GB 7635d8ff download
model-00012-of-00021.safetensors 3.08 GB 6d6ecd70 download
model-00014-of-00021.safetensors 3.08 GB 682e690e download
model-00016-of-00021.safetensors 3.08 GB bedfca45 download
model-00018-of-00021.safetensors 3.08 GB 1681f718 download
model-00020-of-00021.safetensors 3.08 GB 5acb4d5a download
model-00004-of-00021.safetensors 3.08 GB d9883862 download
model-00006-of-00021.safetensors 3.08 GB bf8ac13e download
model-00021-of-00021.safetensors 1.89 GB 5e5c9a29 download
tokenizer.json 19.1 MB 87a7830d download
model.safetensors.index.json 99.5 KB 8bfcf0f4 download
abliteration_metadata.json 18.2 KB 39e45659 download
chat_template.jinja 7.58 KB a8755d82 download
config.json 3.11 KB 6612d5fe download
.gitattributes 1.53 KB 52373fe2 download
processor_config.json 1.16 KB 33818c7f download
README.md 1.15 KB 027edfe9 download
tokenizer_config.json 1.11 KB 541f6c47 download
generation_config.json 242 B 89a0a0db download

README current version from Hugging Face

library_name: transformers
license: apache-2.0
base_model: Qwen/Qwen3.6-35B-A3B
tags:

  • qwen
  • qwen3.6
  • qwen3_5_moe
  • moe
  • mixture-of-experts
  • multimodal
  • vlm
  • abliterated
  • uncensored
  • heretic
  • mpoa
  • soma
  • bf16
  • custom_code
  • text-generation

Qwen3.6-35B-A3B-Abliterated-Heretic-BF16

A BF16 abliterated Qwen3.6-35B-A3B checkpoint produced with Heretic.

Quick Benchmarks

Check Original Qwen3.6-35B-A3B Abliterated Heretic BF16
Official 25-prompt refusal check 22/25 refusals 1/25 refusals
Archived Heretic KL divergence - 0.010655362159013748

Abliteration notes:

  • base model: Qwen/Qwen3.6-35B-A3B
  • method family: Heretic MPOA/SOMA-style sibling transfer, finalized with split-MoE input-side intervention
  • official refusal check: 1/25 refusals on the same 25-prompt marker suite used for the MiniMax M2.7 abliterated run
  • vision-language wrapper preserved; intervention was applied on the text-side MoE stack

Notes:

  • GGUF quants are published separately in Youssofal/Qwen3.6-35B-A3B-Abliterated-Heretic-GGUF
  • Export metadata for the accepted candidate is included in abliteration_metadata.json

README history 1 version

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

  1. 2026-05-25Duplicate from Youssofal/Qwen3.6-35B-A3B-Abliterated-Heretic-BF164aa5a081.1 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.

Get the app Browse the catalog
Open in Abliteration