← back to catalog · registered 2026-09-26 18:57

xCloudinfo/Qwen3.5-9B-Uncensored-GGUF

xCloudinfo Qwen 9B GGUF 262K ctx
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/xCloudinfo%2FQwen3.5-9B-Uncensored-GGUF"
Response includes
  • classification m-uncensored
  • files 9
  • benchmarks 11 entries
  • hub_downloads_all_time 657
  • author_summary 26 models
  • readme_text full
10 credits · hourly refresh · ~4 KB payload Get an API key →
Abliteration classifier · v1.0.0
M-U
Primary method

Uncensored (method unknown)

No other method signals detected in this model.
Confidence
LOW
Why this label 3 signals
Weak or ambiguous signals. Best guess based on catalog patterns; treat as tentative and check the evidence below.
  • 'uncensored' in name/tags but no 'abliterated' marker
  • method not identifiable from author declaration alone
  • may be DPO fine-tune, prompt engineering, or unknown technique
Refusal direction extraction

No specific extraction method could be identified for this model. The producer either did not document it or used a proprietary pipeline.

What is a refusal direction? →
Downloads · lifetime
657
Likes
3
Model age
8w ago
created 2026-09-26
Downloads over time
Now 6.5K → from 0 ↑ 0%
02.4K4.7K7.1K 0 on Sep 26 6.5K on Nov 22 SepOctNov
Sep 26 → Nov 22 · 58 snapshots · spans 57 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 2.4 UGI
Natural Intelligence 17.62 UGI
Political lean -12.2% UGI
Sensitive-Info 14.65 UGI
SocPol 0.9 UGI
UGI 17.27 UGI
Willingness (10) 2.2 UGI
W10-Adherence 1.5 UGI
W10-Direct 3 UGI
Writing 33.52 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.

Metadata

License
apache-2.0
Quantizations
IQ2 IQ4 Q4_K Q5_K Q6_K Q8_0
Tags
gguf abliterated uncensored heretic zh en base_model:Qwen/Qwen3.5-9B base_model:quantized:Qwen/Qwen3.5-9B license:apache-2.0 endpoints_compatible region:us imatrix

Related

Total size
35.2 GB
Files
9
Quantizations
8
Registered
2026-09-26 18:57
Last updated on HF
2026-10-26 05:05

Files by quantization

Q8_0 1 file 8.87 GB
Qwen3.5-9B-Uncensored-xCloud-Q8_0.gguf 8.87 GB 7143651e download
Q6_K 1 file 6.85 GB
Qwen3.5-9B-Uncensored-xCloud-Q6_K.gguf 6.85 GB 3fc796cd download
Q5_K 1 file 6.02 GB
Qwen3.5-9B-Uncensored-xCloud-Q5_K_M.gguf 6.02 GB 07e80370 download
Q4_K 1 file 5.24 GB
Qwen3.5-9B-Uncensored-xCloud-Q4_K_M.gguf 5.24 GB 1087cdb4 download
IQ4 1 file 4.84 GB
Qwen3.5-9B-Uncensored-xCloud-IQ4_XS.gguf 4.84 GB a0e8097f download
IQ2 1 file 3.36 GB
Qwen3.5-9B-Uncensored-xCloud-IQ2_M.gguf 3.36 GB 92f10448 download
F16 1 file 879 MB
mmproj-Qwen3.5-9B-Uncensored-xCloud-f16.gguf 879 MB 4131f430 download
Auxiliary files 2 files 4.23 KB
README.md 2.22 KB b616c315 download
.gitattributes 2.01 KB 4cf3ad6b download

README current version from Hugging Face


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

  • gguf
  • abliterated
  • uncensored
  • heretic

Qwen3.5-9B-Uncensored-GGUF

云碩(xCloudinfo)自製的「降低過度拒絕」版本,底模 Qwen/Qwen3.5-9B。
與我方 Qwen3.8-27B-Uncensored-GGUF 同一套方法論,
補齊小顯存機器(單卡 8–12GB、邊緣機)可用的小尺寸選項。

方法

用 Heretic 自動搜參(200 trials,同時最小化拒答率與 KL 散度守住原能力),
迭代兩輪:iter1 收斂後,對其成品再跑一輪 iter2 進一步下壓。

誠實揭露的數字(不同尺、不同結果)

  • safetensors 合併模型(greedy 解碼):我方 10 題嚴重危害 held-out,iter1 拒答 3/10 → iter2 拒答 0/10,
    無害題 coherence 完好(繁中流暢正確)。
  • 量化後 GGUF、經 llama-server 實際載入測(預設取樣,非 greedy):Q6_K 與 Q4_K_M 皆實測 拒答 2/10。
    量化與取樣方式的變動會讓拒答率有小幅回升,這裡如實列出兩種情境的數字,不只挑好看的那個。

內含檔案

  • Qwen3.5-9B-Uncensored-xCloud-{Q8_0,Q6_K,Q5_K_M,Q4_K_M,IQ4_XS,IQ2_M}.gguf —— 六階量化
  • mmproj-Qwen3.5-9B-Uncensored-xCloud-f16.gguf —— 視覺塔(多模態)

技術修正

轉檔時遇到 Qwen3.5 系列共通的「幽靈 MTP」metadata 問題:qwen35.block_count 比實際層數多算一層
(本例 33 vs 實際 32 層),nextn_predict_layers 誤標為 1。已修正兩個 metadata 欄位再進行量化,
避免載入端因層數對不上而出錯或行為異常。

使用方式

llama.cpp / llama-server / 任何相容 GGUF 的推論引擎皆可直接載入。範例:

llama-server -m Qwen3.5-9B-Uncensored-xCloud-Q4_K_M.gguf --jinja \
  --chat-template-kwargs '{"enable_thinking":false}'

授權與使用前提

繼承 Qwen3.5-9B 原廠授權(Apache 2.0)。
本版本移除的是對合法、邊界性題目的反射式拒答,不代表對任何非法用途的背書;使用者須自行承擔合法使用之責任。

README history 2 versions

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

  1. 2026-10-26 Upload README.md with huggingface_hub ff41057 5.2 KB
    Loading...
  2. 2026-09-26 Upload README.md with huggingface_hub 81784e0 2.2 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