← back to catalog · registered 2026-10-12 00:59

xohaohe/Qwen3.8-27B-TURBO-Fable5.5-Cold-Fusion-L-Uncensored-fp16-oQ4e-mtp

xohaohe 27B multimodal
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/xohaohe%2FQwen3.8-27B-TURBO-Fable5.5-Cold-Fusion-L-Uncensored-fp16-oQ4e-mtp"
Response includes
  • classification m-uncensored
  • files 15
  • author_summary 1 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 · 30-day
0
Likes
0
Model age
today
created 2026-10-12

Metadata

License
apache-2.0
Tags
mlx safetensors qwen3_5 omlx oq mixed-precision multimodal mtp image-text-to-text conversational license:apache-2.0 4-bit
Total size
16.7 GB
Files
15
Quantizations
1
Registered
2026-10-12 00:59
Last updated on HF
2026-10-12 00:37

Files by quantization

Auxiliary files 15 files 16.7 GB
model-00002-of-00004.safetensors 4.70 GB 1ffe2164 download
model-00003-of-00004.safetensors 4.69 GB 9c8d609e download
model-00001-of-00004.safetensors 4.66 GB 1b0a5bf8 download
model-00004-of-00004.safetensors 2.67 GB d7af5297 download
tokenizer.json 12.2 MB 0997f410 download
vocab.json 6.41 MB 0aa0ce06 download
model.safetensors.index.json 207 KB 4f7d8b32 download
config.json 37.9 KB 5d73cb4b download
oq_imatrix_report.json 30.4 KB 2a4ef46c download
chat_template.jinja 16.5 KB fa46c180 download
tokenizer_config.json 8.48 KB 1a7e62d0 download
README.md 1.85 KB abacb271 download
.gitattributes 1.53 KB 52373fe2 download
preprocessor_config.json 390 B 2ea84a43 download
generation_config.json 202 B 023756cf download

README current version from Hugging Face


library_name: mlx
license: apache-2.0
pipeline_tag: image-text-to-text
tags:

  • mlx
  • omlx
  • oq
  • mixed-precision
  • multimodal
  • mtp

Qwen3.8-27B-TURBO-Fable5.5-Cold-Fusion-L-Uncensored-fp16-oQ4e-mtp

由 xohaohe(华哥)发布的 27B 级 MLX 多模态混合精度模型,开放下载。

发布者介绍

发布者说明:使用 oMLX 0.7.1 进行全面混合精度量化处理,保留视觉模块和 MTP。发布者个人评价为“目前最优秀的小模型”;尚待补充可复现对比成绩。

技术信息

  • 稠密 Qwen 架构,64 层语言网络,隐藏维度 5120。
  • 默认 4-bit affine,group size 64;130 个模块配置为 5-bit、2 个为 6-bit。
  • 权重含量化打包整数、FP16 和 FP32;视觉编码器权重保留 FP32。
  • 保留 27 层视觉编码器、图片/视频模板和预处理配置;具体多模态支持依赖引擎。
  • 保留 1 层 MTP;是否加速取决于引擎支持和预测接受率。
  • 配置上下文:262,144 tokens(256K),未验证满长度质量。
  • 4 个 Safetensors 分片,共 17.94 GB(16.71 GiB)。
  • 校准报告:oqe_code_multilingual,128 个样本,每个 512 tokens。

测试成绩

发布者本量化版本的测试成绩及测试设置待补充。上游 MXFP4/MXFP8 成绩不作为本版本实测成绩。

量化作者

本量化版本的一手作者:xohaohe(华哥)。根据作者提供的制作信息,本版本使用 oMLX 0.7.1 进行全面混合精度量化,保留多模态视觉模块与 MTP 权重,面向 Apple Silicon 发布。

下载

hf download xohaohe/Qwen3.8-27B-TURBO-Fable5.5-Cold-Fusion-L-Uncensored-fp16-oQ4e-mtp --local-dir ./Qwen3.8-27B-TURBO-Fable5.5-Cold-Fusion-L-Uncensored-fp16-oQ4e-mtp

使用支持该架构的 MLX 多模态引擎加载完整目录。默认生成配置:temperature=1.0、top_p=0.95、top_k=20。

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