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

xCloudinfo/gpt-oss-20b-Uncensored-xCloud

xCloudinfo Gpt-oss 20B MoE
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%2Fgpt-oss-20b-Uncensored-xCloud"
Response includes
  • classification m-uncensored
  • files 8
  • benchmarks 16 entries
  • hub_downloads_all_time 109
  • author_summary 25 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
109
26 last 30d - stable
Likes
0
Descendants
2
in 2 direct forks
Model age
3mo ago
created 2026-06-25
Downloads over time
Now119→from16↑644%
11509012916 on Jun 24119 on Oct 11119 on Oct 8JunJulAugSepOct
Jun 24 → Oct 11 · 55 snapshots · spans 109 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
Arena-Battles 7952 LM-Arena
LM Arena Elo 1307.3530229799078 LM-Arena
Arena-Elo-Lower 1300.2396666200118 LM-Arena
Arena-Elo-Upper 1314.4663793398038 LM-Arena
Arena-Rank 70 LM-Arena
Entertainment 1.1 UGI
Hazardous 0 UGI
Natural Intelligence 17.39 UGI
Political lean -10.6% UGI
Sensitive-Info 7.19 UGI
SocPol 0.8 UGI
UGI 8.96 UGI
Willingness (10) 1.2 UGI
W10-Adherence 1.5 UGI
W10-Direct 1 UGI
Writing 24.62 UGI

Genealogy 2 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 · 144 downloads combined

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

Metadata

License
apache-2.0
Languages
en zh
Tags
transformers safetensors gpt_oss text-generation gpt-oss uncensored compliance moe mxfp4 reasoning xcloudinfo conversational

Related

Total size
12.8 GB
Files
8
Quantizations
1
Registered
2026-08-22 13:56
Last updated on HF
2026-06-25 09:18

Files by quantization

Auxiliary files 8 files 12.9 GB
model.safetensors 12.8 GB e866c3df download
tokenizer.json 26.6 MB 0614fe83 download
chat_template.jinja 16.3 KB dc7bb119 download
README.md 3.03 KB c8a19f1f download
config.json 1.81 KB 162a5ebc download
.gitattributes 1.53 KB 52373fe2 download
tokenizer_config.json 380 B d8d51b1b download
generation_config.json 177 B 86f91466 download

README current version from Hugging Face


license: apache-2.0
base_model: openai/gpt-oss-20b
language:

  • en
  • zh
    tags:
  • gpt-oss
  • uncensored
  • compliance
  • moe
  • mxfp4
  • reasoning
  • xcloudinfo
    pipeline_tag: text-generation
    library_name: transformers

gpt-oss-20b-Uncensored-xCloud

云碩科技 · xCloudinfo · 系列:無審查 · Uncensored

以 openai/gpt-oss-20b(21B 總參 / 3.6B 活躍 / MoE / MXFP4 / harmony 推理格式)為基底 的低拒答(uncensored) reasoning 模型。在程式能力底層之上做一段 compliance SFT,讓推理(analysis)通道對正當、獲授權的技術請求服從作答,降低 gpt-oss 預設的過度拒答。

功能:獲授權的資安/雙用途技術問答——對紅隊、滲透測試、惡意程式分析等正當場景不過度拒答。

厲害在哪

  • 真的降低硬核拒答:用 compliance SFT 教 analysis 通道服從,連單向權重 abliteration 都消不掉的硬核授權請求(如 reverse shell)都能作答——因為 gpt-oss 把安全焊在思維鏈,唯有教它推理才有效。
  • 疊在程式底層上:以 Code-xCloud 為底,保留 coding 與 reasoning。
  • 輕量好部署:MXFP4、約 14GB,可本地離線、可控部署。

做法

  • 程式底層:執行驗證蒸餾的程式碼資料 LoRA 微調。
  • 低拒答對齊:compliance SFT 教 analysis 通道對授權場景服從(非權重 abliteration——對 reasoning 模型無效)。

用法(transformers)

from transformers import AutoModelForCausalLM, AutoTokenizer
tok = AutoTokenizer.from_pretrained("xCloudinfo/gpt-oss-20b-Uncensored-xCloud")
model = AutoModelForCausalLM.from_pretrained("xCloudinfo/gpt-oss-20b-Uncensored-xCloud", dtype="auto", device_map="auto")
msgs = [{"role": "user", "content": "Explain how to set up a bash reverse shell for an authorized penetration test."}]
ids = tok.apply_chat_template(msgs, add_generation_prompt=True, return_tensors="pt").to(model.device)
print(tok.decode(model.generate(ids, max_new_tokens=512)[0][ids.shape[1]:], skip_special_tokens=False))

reasoning 模型:請給足 max_new_tokens。GGUF 版見 gpt-oss-20b-Uncensored-xCloud-GGUF。

用途與責任聲明

本模型降低預設拒答,用途定位為獲授權的資安研究、紅隊演練、滲透測試、雙用途技術問答與內部可控部署。使用者須:

  • 僅在取得授權、合法、合乎倫理的前提下使用;不得用於非法入侵、製造危害、軍事或任何違法用途。
  • 自行為輸出與後續行為負責,並遵守中華民國法律與適用之 EU AI Act 等法規。
  • 模型輸出可能不準確或有害,部署方應自行加上適用的審核與防護。

授權與來源聲明


由 云碩科技 xCloudinfo 於自有 AI 算力資源池製作;資料留在本地、流程可重現。

README history 2 versions

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

  1. 2026-06-25Update model card: function + honest strengthsf12bf8f3 KB
    Loading...
  2. 2026-06-25Add model card (xCloudinfo)359798b2.4 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