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

xCloudinfo/Qwen3.8-27B-Uncensored-GGUF

xCloudinfo Qwen 27B GGUF second-order 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%2FQwen3.8-27B-Uncensored-GGUF"
Response includes
  • classification m8
  • files 9
  • hub_downloads_all_time 4,066
  • author_summary 25 models
  • readme_text full
10 credits · hourly refresh · ~4 KB payload Get an API key →
Abliteration classifier · v1.0.0
M8
Primary method

Repackaging (quantization)

Applied on top of direct removal inherited from the base model.
Confidence
MEDIUM
Inherited from base model
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=1
  • assume M1 (base ablation) + M8 (GGUF quant) - default when producer unknown
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
4K
Likes
1
Model age
7w ago
created 2026-08-17
Downloads over time
Now5.6K→from389↑1,338%
1292.1K4.1K6.1K389 on Aug 175.6K on Oct 11AugSepOct
Aug 17 → Oct 11 · 49 snapshots · spans 55 days

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
zh en
Quantizations
IQ2 IQ4 Q4_K Q5_K Q6_K Q8_0
Tags
gguf uncensored abliterated qwen3.5 reasoning text-generation zh en base_model:Qwen/Qwen3.8-27B base_model:quantized:Qwen/Qwen3.8-27B license:apache-2.0 endpoints_compatible

Related

Total size
106 GB
Files
9
Quantizations
7
Registered
2026-08-22 13:56
Last updated on HF
2026-09-16 01:29

Files by quantization

Q8_0 1 file 27.1 GB
Qwen3.8-27B-Uncensored-xCloud-Q8_0.gguf 27.1 GB 00f87fb5 download
Q6_K 1 file 20.9 GB
Qwen3.8-27B-Uncensored-xCloud-Q6_K.gguf 20.9 GB 19700ec0 download
Q5_K 1 file 18.2 GB
Qwen3.8-27B-Uncensored-xCloud-Q5_K_M.gguf 18.2 GB f6a43619 download
Q4_K 1 file 15.7 GB
Qwen3.8-27B-Uncensored-xCloud-Q4_K_M.gguf 15.7 GB 46f29f8c download
IQ4 1 file 14.3 GB
Qwen3.8-27B-Uncensored-xCloud-IQ4_XS.gguf 14.3 GB 6e9705d8 download
IQ2 1 file 9.64 GB
Qwen3.8-27B-Uncensored-xCloud-IQ2_M.gguf 9.64 GB 147feffc download
Auxiliary files 3 files 13.0 MB
Qwen3.8-27B-Uncensored-xCloud-imatrix.dat 13.0 MB c51f6ac2 download
.gitattributes 2.01 KB f4839bb5 download
README.md 1.85 KB b2ca5be3 download

README current version from Hugging Face


license: apache-2.0
base_model: xCloudinfo/Qwen3.8-27B-Uncensored-xCloud
base_model_relation: quantized
language:

  • zh
  • en
    pipeline_tag: text-generation
    library_name: llama.cpp
    tags:
  • gguf
  • uncensored
  • abliterated
  • qwen3.5
  • imatrix
  • xCloudinfo

Qwen3.8-27B-Uncensored-xCloud — GGUF

xCloudinfo/Qwen3.8-27B-Uncensored-xCloud
的 GGUF 量化版,供 llama.cpp/agent 調度使用。由 云碩科技(xCloudinfo) 自行轉檔量化。

底模去除拒絕方向(abliteration)的方法與行為,見上游 safetensors 版的說明。

量化版本

檔案 位元 大小 用途
...-Q8_0.gguf 8-bit ~28 GB 近乎無損,品質最高
...-Q6_K.gguf 6-bit ~21 GB 高品質
...-Q5_K_M.gguf 5-bit ~19 GB 品質與體積的均衡
...-Q4_K_M.gguf 4-bit ~17 GB agent 調度主力,單張 24 GB 卡即可
...-IQ4_XS.gguf 4-bit ~15 GB 更小的 4-bit(imatrix)
...-IQ2_M.gguf 2-bit ~10 GB 低 VRAM/邊緣(imatrix,品質較低)

...-imatrix.dat 為重要度矩陣(importance matrix),用於 IQ 系列量化。
K 系列(Q4_K_M 等)亦套用同一份 imatrix 以提升品質。

執行(llama.cpp)

# OpenAI 相容伺服器,供 agent 調度
llama-server -m Qwen3.8-27B-Uncensored-xCloud-Q4_K_M.gguf \
  -ngl 99 -c 8192 --host 0.0.0.0 --port 8080 --jinja

-ngl 99 將所有層放到 GPU。這是混合(SSM/線性注意力)模型,llama.cpp 會在單一裝置上正確處理其遞迴狀態。
需使用支援 qwen35 架構的 llama.cpp 版本(近期 master 已內建)。

負責任使用

移除拒絕方向等於移除一層安全機制。使用者需自行負責這些權重的用途,並遵守 Apache-2.0 授權與所有適用法律。

README history 4 versions

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

  1. 2026-09-16Upload README.md with huggingface_hub5207e653.5 KB
    Loading...
  2. 2026-09-14Upload README.md with huggingface_hub03b35393.1 KB
    Loading...
  3. 2026-08-24Upload folder using huggingface_hubb4a4d5e2.3 KB
    Loading...
  4. 2026-08-17GGUF quants Q8_0..IQ4_XS + imatrix99fe92f1.9 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