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

xCloudinfo/Gemma-4-26B-A4B-TAIDE-zhTW-abliterated-GGUF

xCloudinfo Gemma 26B GGUF MoE multimodal 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%2FGemma-4-26B-A4B-TAIDE-zhTW-abliterated-GGUF"
Response includes
  • classification m8
  • files 4
  • hub_downloads_all_time 4,265
  • 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
495 last 30d - stable
Likes
0
Model age
4mo ago
created 2026-06-09
Downloads over time
Now4.4K→from145↑2,906%
01.6K3.2K4.8K145 on Jun 104.4K on Oct 11JunJulAugSepOct
Jun 10 → Oct 11 · 57 snapshots · spans 123 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
Quantizations
Q4_K
Tags
gguf gemma4 taide traditional-chinese zh-tw llama-cpp multimodal vision mmproj abliterated uncensored xcloudinfo

Related

Total size
15.6 GB
Files
4
Quantizations
3
Registered
2026-08-22 13:56
Last updated on HF
2026-08-24 12:07

Files by quantization

Q4_K 1 file 15.6 GB
model-Q4_K_M.gguf 15.6 GB 83472d7a download
F16 1 file 1.11 GB
mmproj-Gemma-4-26B-A4B-TAIDE-zhTW-abliterated-f16.gguf 1.11 GB 2822805e download
Auxiliary files 2 files 3.85 KB
README.md 2.23 KB 9c942657 download
.gitattributes 1.63 KB 8b30d3dc download

README current version from Hugging Face


license: apache-2.0
base_model: xCloudinfo/Gemma-4-26B-A4B-TAIDE-zhTW-abliterated
language:

  • zh
    tags:
  • gemma4
  • taide
  • traditional-chinese
  • zh-tw
  • gguf
  • llama-cpp
  • multimodal
  • vision
  • mmproj
  • abliterated
  • uncensored
  • xcloudinfo
    pipeline_tag: image-text-to-text

Gemma-4-26B-A4B-TAIDE-zhTW-abliterated-GGUF

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

⚠️ 經 abliteration 處理,會比原版更少拒答。請僅在你被授權的範圍內使用。

xCloudinfo/Gemma-4-26B-A4B-TAIDE-zhTW-abliterated 的 GGUF,含視覺投影器(mmproj),可圖文對話。

檔案 用途
…-Q4_K_M.gguf 語言模型(文字)
mmproj-Gemma-4-26B-A4B-TAIDE-zhTW-f16.gguf 視覺投影器(vision)

視覺塔在 abliteration 中未更動,故此 mmproj 與原版相同、完全適用。

用法(含圖文)

llama-server -m Gemma-4-26B-A4B-TAIDE-zhTW-abliterated-Q4_K_M.gguf \
  --mmproj mmproj-Gemma-4-26B-A4B-TAIDE-zhTW-f16.gguf -c 4096 -ngl 99

純文字可省略 --mmproj。

原理

依 Arditi et al. (2024),以權重正交化移除殘差流的拒絕方向(61 矩陣、strength 1.0),不重訓、僅作用語言模型。詳見 merged 版模型卡。

用途與責任

  • 設計用途:降低過度拒絕,供授權範圍內研究與應用。
  • 已移除安全拒絕傾向,使用者須自行加上安全防護與輸出審查,並對用途與後果負完全責任;不得用於有害或違法用途。

授權與來源聲明


由 云碩科技 xCloudinfo 於自有 AI 算力資源池製作。

README history 5 versions

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

  1. 2026-08-24Upload folder using huggingface_huba8e93422.9 KB
    Loading...
  2. 2026-06-15Upload folder using huggingface_hubd066dde2.2 KB
    Loading...
  3. 2026-06-14Upload folder using huggingface_huba041cb72.2 KB
    Loading...
  4. 2026-06-10Document vision (mmproj) usage2e934c73.4 KB
    Loading...
  5. 2026-06-10Upload folder using huggingface_hube26a1da3 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