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

ryo559/Qwen3-8B-JP-Uncensored

ryo559 Qwen 8.2B
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/ryo559%2FQwen3-8B-JP-Uncensored"
Response includes
  • classification m1
  • files 8
  • benchmarks 11 entries
  • hub_downloads_all_time 322
  • author_summary 2 models
  • readme_text full
10 credits · hourly refresh · ~4 KB payload Get an API key →
Abliteration classifier · v1.0.0
M1
Primary method

Direct removal

No other method signals detected in this model.
Confidence
MEDIUM
Why this label 3 signals
Method inferred from partial signals - repository name, related files, or tag patterns. Producer identity not confirmed; label may sharpen or shift as we gather more evidence.
  • 'abliterated' in name/tags
  • is_gguf=0 (base model)
  • no specific method indicators - defaulting to M1 (most common)
Refusal direction extracted via
Extraction technique

Difference-of-means

Confidence
MEDIUM
Why we say so
primary_method=M1; difference-of-means is the reference extraction for M1/M3 (Arditi 2024)
Downloads · lifetime
322
213 last 30d - active
Likes
0
Descendants
3
in 3 direct forks
Model age
7mo ago
created 2026-03-03
Downloads over time
Now413→from48↑760%
3017031045048 on Mar 4413 on Oct 11MarAprMayJunJulAugSepOct
Mar 4 → Oct 11 · 71 snapshots · spans 221 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 2.9 UGI
Natural Intelligence 15.15 UGI
Political lean -9.9% UGI
Sensitive-Info 18.27 UGI
SocPol 1.4 UGI
UGI 32.18 UGI
Willingness (10) 6 UGI
W10-Adherence 7 UGI
W10-Direct 5 UGI
Writing 27.96 UGI

Genealogy 3 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.

Variants by this author 2 formats · 1K downloads combined

The same weights this author released in different packaging. Pick the format that matches your runtime.

Metadata

License
apache-2.0
Languages
ja en
Tags
safetensors qwen3 abliterated uncensored japanese ja en base_model:Qwen/Qwen3-8B base_model:finetune:Qwen/Qwen3-8B license:apache-2.0 region:us

Related

Total size
15.3 GB
Files
8
Quantizations
1
Registered
2026-08-22 13:56
Last updated on HF
2026-03-03 09:16

Files by quantization

Auxiliary files 8 files 15.3 GB
model.safetensors 15.3 GB 02671e90 download
tokenizer.json 10.9 MB be756060 download
chat_template.jinja 4.07 KB 01be9b30 download
config.json 1.56 KB 7a5df4a4 download
.gitattributes 1.53 KB 52373fe2 download
README.md 1.09 KB ce685ff8 download
tokenizer_config.json 664 B 9fd0fb48 download
generation_config.json 213 B 57bd7165 download

README current version from Hugging Face


language:

  • ja
  • en
    license: apache-2.0
    base_model: Qwen/Qwen3-8B
    tags:
  • abliterated
  • uncensored
  • japanese
  • qwen3

Qwen3-8B-JP-Uncensored

Qwen3-8Bから拒否方向ベクトルを外科的に除去(Abliteration)したモデル。

特徴

  • 日本語+英語の拒否パターンを両方除去(各20件、計40件のプロンプトセット)
  • Norm-Preserving Abliteration: 能力劣化を最小限に抑制
  • ベースモデルの日本語性能をそのまま維持

Abliteration詳細

  • 最適レイヤー: 34 (score: 207.9685)
  • 修正レイヤー: [31, 32, 33, 34]
  • 修正重み行列数: 8
  • テスト結果: 8/8 拒否なし

使い方

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("ryo559/Qwen3-8B-JP-Uncensored", torch_dtype="auto", device_map="auto")
tokenizer = AutoTokenizer.from_pretrained("ryo559/Qwen3-8B-JP-Uncensored")

免責事項

このモデルは研究・教育目的で公開しています。生成されるコンテンツの責任はユーザーにあります。

README history 3 versions

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

  1. 2026-03-03Full uncensored: multi-direction abliteration + DPO0ff03331.1 KB
    Loading...
  2. 2026-03-03Add model card024babc1.1 KB
    Loading...
  3. 2026-03-03Abliterated Qwen3-8B (JP+EN, norm-preserving)55039e75.1 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.

Open in Abliteration