玩模型开放模型 · 清晰来源 · 自主选择
返回模型目录
HauhauCS /

Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive

HF 源站

HauhauCS/Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive is an uncensored open-source model hosted on Hugging Face.

Image-Text-to-TextGGUFChineseEnglishMultilingualLicense: apache-2.035Buncensoredqwen3.6moevisionmultimodalimatrixconversational
README.md说明 / 使用方法

license: apache-2.0 tags:

  • uncensored
  • qwen3.6
  • moe
  • gguf
  • vision
  • multimodal language:
  • en
  • zh
  • multilingual pipeline_tag: image-text-to-text base_model: Qwen/Qwen3.6-35B-A3B

Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive

加入 Discord 获取更新、路线图、项目信息,或进行交流探讨。

由 HauhauCS 进行去审查(uncensored)的 Qwen3.6-35B-A3B。0/465 拒绝率。

HuggingFace 的“硬件兼容性(Hardware Compatibility)”组件无法识别 K_P 量化版本 —— 它显示的文件数量可能少于实际存在的文件。请点击 "View +X variants" 或进入 Files and versions 页面查看所有可供下载的文件。

关于模型 (About)

未对数据集或模型能力做任何缩减更改。功能完全保留,100% 达成原作者的设计意图 —— 只是去除了拒绝回答(refusals)的行为。

本系列旨在成为市面上最顶级的无损去审查模型。

激进版本 (Aggressive Variant)

更强力的去审查 —— 模型已完全解锁,绝不拒绝任何提示词。可能会偶尔附带简短的免责声明(这是固化在基座模型预训练中的内容,而非拒绝回答),但完整内容始终会完整生成。

如果你需要保留部分安全护栏的更保守的去审查版本,可在发布后查看 Balanced(平衡)版本。

下载列表 (Downloads)

文件 (File)量化格式 (Quant)BPW大小 (Size)
Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive-Q8_K_P.ggufQ8_K_P10.0644 GB
—Q8_08.5—
Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive-Q6_K_P.ggufQ6_K_P7.0731 GB
—Q6_K6.6—
Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive-Q5_K_P.ggufQ5_K_P6.4728 GB
—Q5_K_M5.7—
Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive-Q4_K_P.ggufQ4_K_P5.4023 GB
Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive-Q4_K_M.ggufQ4_K_M4.8821 GB
Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive-IQ4_NL.ggufIQ4_NL4.5620 GB
Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive-IQ4_XS.ggufIQ4_XS4.3219 GB
Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive-Q3_K_P.ggufQ3_K_P4.3919 GB
—Q3_K_M3.9—
Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive-IQ3_M.ggufIQ3_M3.5615 GB
Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive-Q2_K_P.ggufQ2_K_P3.4615 GB
Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive-IQ2_M.ggufIQ2_M2.6911 GB
mmproj-Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive-f16.ggufmmproj (f16)—899 MB

所有量化版本均基于重要性矩阵(imatrix)生成,以便在去拒绝(abliterated)权重上达到最佳的模型质量保持。

什么是 K_P 量化?(What are K_P quants?)

K_P("Perfect")量化是 HauhauCS 的定制量化技术,利用特定模型的专属分析,在最关键的层和权重上有针对性地保留质量。每个模型都拥有量身定制的专属优化量化配置文件。

与基础量化相比,K_P 量化仅增加约 5-15% 的文件大小,却能有效将生成质量提升 1 到 2 个量化档位。完全兼容 llama.cpp、LM Studio 以及任何兼容 GGUF 的运行时环境 —— 无需特殊定制编译版本。

注意: K_P 量化在 LM Studio 的量化列中可能会显示为“?”。这仅仅是显示问题 —— 模型可以正常加载并流畅运行。

模型规格 (Specs)

  • 总参数量 35B,每次前向传播激活约 3B 参数 (MoE)
  • 256 个专家 (Experts),每个 Token 路由激活 8 个专家
  • 混合架构:线性注意力 (Linear Attention) + 全 Softmax 注意力 (Full Softmax Attention)(比例为 3:1)
  • 40 层
  • 原生支持 262K 上下文
  • 原生多模态支持(文本、图像、视频)
  • 基于 Qwen/Qwen3.6-35B-A3B

推荐参数设置 (Recommended Settings)

来自 Qwen 官方团队的推荐配置:

思考模式(Thinking mode,默认):

  • 通用任务:temperature=1.0, top_p=0.95, top_k=20, min_p=0, presence_penalty=1.5
  • 编码 / 精确任务:temperature=0.6, top_p=0.95, top_k=20, min_p=0, presence_penalty=0

非思考模式(Non-thinking mode):

  • 通用任务:temperature=0.7, top_p=0.8, top_k=20, min_p=0, presence_penalty=1.5
  • 推理任务:temperature=1.0, top_p=1.0, top_k=40, min_p=0, presence_penalty=2.0

重要事项:

  • 保持至少 128K 上下文以完整保留思考推理能力
  • 在 llama.cpp 中使用 --jinja 参数以确保正确处理对话模板 (chat template)
  • 视觉理解能力需要配合 mmproj 文件与主 GGUF 文件一同加载

使用方法 (Usage)

适用于 llama.cpp、LM Studio、Jan、koboldcpp 及其他兼容 GGUF 的运行时。

llama-cli -m Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive-Q4_K_P.gguf \
  --mmproj mmproj-Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive-f16.gguf \
  --jinja -c 131072 -ngl 99

其他模型 (Other Models)

  • HauhauCS 的 HuggingFace 主页
源站原文

license: apache-2.0 tags:

  • uncensored
  • qwen3.6
  • moe
  • gguf
  • vision
  • multimodal language:
  • en
  • zh
  • multilingual pipeline_tag: image-text-to-text base_model: Qwen/Qwen3.6-35B-A3B

Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive

Join the Discord for updates, roadmaps, projects, or just to chat.

Qwen3.6-35B-A3B uncensored by HauhauCS. 0/465 Refusals.

HuggingFace's "Hardware Compatibility" widget doesn't recognize K_P quants — it may show fewer files than actually exist. Click "View +X variants" or go to Files and versions to see all available downloads.

About

No changes to datasets or capabilities. Fully functional, 100% of what the original authors intended - just without the refusals.

These are meant to be the best lossless uncensored models out there.

Aggressive Variant

Stronger uncensoring — model is fully unlocked and won't refuse prompts. May occasionally append short disclaimers (baked into base model training, not refusals) but full content is always generated.

For a more conservative uncensor that keeps some safety guardrails, check the Balanced variant when it's available.

Downloads

FileQuantBPWSize
Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive-Q8_K_P.ggufQ8_K_P10.0644 GB
—Q8_08.5—
Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive-Q6_K_P.ggufQ6_K_P7.0731 GB
—Q6_K6.6—
Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive-Q5_K_P.ggufQ5_K_P6.4728 GB
—Q5_K_M5.7—
Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive-Q4_K_P.ggufQ4_K_P5.4023 GB
Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive-Q4_K_M.ggufQ4_K_M4.8821 GB
Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive-IQ4_NL.ggufIQ4_NL4.5620 GB
Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive-IQ4_XS.ggufIQ4_XS4.3219 GB
Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive-Q3_K_P.ggufQ3_K_P4.3919 GB
—Q3_K_M3.9—
Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive-IQ3_M.ggufIQ3_M3.5615 GB
Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive-Q2_K_P.ggufQ2_K_P3.4615 GB
Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive-IQ2_M.ggufIQ2_M2.6911 GB
mmproj-Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive-f16.ggufmmproj (f16)—899 MB

All quants generated with importance matrix (imatrix) for optimal quality preservation on abliterated weights.

What are K_P quants?

K_P ("Perfect") quants are HauhauCS custom quantizations that use model-specific analysis to selectively preserve quality where it matters most. Each model gets its own optimized quantization profile.

A K_P quant effectively bumps quality up by 1-2 quant levels at only ~5-15% larger file size than the base quant. Fully compatible with llama.cpp, LM Studio, and any GGUF-compatible runtime — no special builds needed.

Note: K_P quants may show as "?" in LM Studio's quant column. This is a display issue only — the model loads and runs fine.

Specs

  • 35B total parameters, ~3B active per forward pass (MoE)
  • 256 experts, 8 routed per token
  • Hybrid architecture: linear attention + full softmax attention (3:1 ratio)
  • 40 layers
  • 262K native context
  • Natively multimodal (text, image, video)
  • Based on Qwen/Qwen3.6-35B-A3B

Recommended Settings

From the official Qwen authors:

Thinking mode (default):

  • General: temperature=1.0, top_p=0.95, top_k=20, min_p=0, presence_penalty=1.5
  • Coding/precise tasks: temperature=0.6, top_p=0.95, top_k=20, min_p=0, presence_penalty=0

Non-thinking mode:

  • General: temperature=0.7, top_p=0.8, top_k=20, min_p=0, presence_penalty=1.5
  • Reasoning tasks: temperature=1.0, top_p=1.0, top_k=40, min_p=0, presence_penalty=2.0

Important:

  • Keep at least 128K context to preserve thinking capabilities
  • Use --jinja flag with llama.cpp for proper chat template handling
  • Vision support requires the mmproj file alongside the main GGUF

Usage

Works with llama.cpp, LM Studio, Jan, koboldcpp, and other GGUF-compatible runtimes.

llama-cli -m Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive-Q4_K_P.gguf \
  --mmproj mmproj-Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive-f16.gguf \
  --jinja -c 131072 -ngl 99

Other Models

  • HauhauCS on HuggingFace
HUGGING FACE

上游数据

HF 下载量106.9万
HF 点赞量3695

HF 下载量为源站最近 30 天统计,与本站点击分开。

SITE STATS

本站统计

下载点击—
请求数—
匿名去重—最近 7 天—

本站下载量记录按钮点击,不代表下载完成;短时连点会去重。

来源与版本

作者
HauhauCS
最近更新
2026/04/17
最新创建
2026/04/17
HF 趋势
16
数据快照
2026/09/16
版本
未知
许可证
apache-2.0