GCSA-AiLab/Qwen3.8-27B-Abliterated-MTP-GGUF overview
Qwen3.8 27B Abliterated MTP GGUF Unofficial community derivative of Qwen/Qwen3.8 27B https://huggingface.co/Qwen/Qwen3.8 27B , published by GCSA AiLab. 这是 GCSA…
Runs locally from ~480.5 MB disk (4 GB VRAM class GPUs with llama.cpp / guIDE).
Repository Files & Downloads
| File | Type | Quantization | Size | Link |
|---|---|---|---|---|
| Q4_K_M/Qwen3.8-27B-Q4_K_M.gguf | GGUF | Q4_K_M | 15.41 GB | Download |
| Q4_K_M/mmproj-Qwen3.8-27B-Q4_K_M.gguf | GGUF | Q4_K_M | 498.1 MB | Download |
| Q4_K_M/mtp-Qwen3.8-27B-BF16.gguf | GGUF | Q4_K_M | 5.54 GB | Download |
| Q4_K_S/Qwen3.8-27B-Q4_K_S.gguf | GGUF | Q4_K_S | 14.52 GB | Download |
| Q4_K_S/mmproj-Qwen3.8-27B-Q4_K_S.gguf | GGUF | Q4_K_S | 480.5 MB | Download |
| Q4_K_S/mtp-Qwen3.8-27B-BF16.gguf | GGUF | Q4_K_S | 5.54 GB | Download |
| Q6_K/Qwen3.8-27B-Q6_K.gguf | GGUF | Q6_K | 20.57 GB | Download |
| Q6_K/mmproj-Qwen3.8-27B-Q6_K.gguf | GGUF | Q6_K | 589.7 MB | Download |
| Q6_K/mtp-Qwen3.8-27B-BF16.gguf | GGUF | Q6_K | 5.54 GB | Download |
| Q8_0/Qwen3.8-27B-Q8_0.gguf | GGUF | Q8_0 | 26.63 GB | Download |
| Q8_0/mmproj-Qwen3.8-27B-Q8_0.gguf | GGUF | Q8_0 | 600.1 MB | Download |
| Q8_0/mtp-Qwen3.8-27B-BF16.gguf | GGUF | Q8_0 | 5.54 GB | Download |
Model Details
| Model ID | GCSA-AiLab/Qwen3.8-27B-Abliterated-MTP-GGUF |
|---|---|
| Author | GCSA-AiLab |
| Pipeline | image-text-to-text |
| License | apache-2.0 |
| Base model | Qwen/Qwen3.8-27B |
| Last modified | 2026-10-09T12:10:24.000Z |
Model README
---
license: apache-2.0
base_model: Qwen/Qwen3.8-27B
pipeline_tag: image-text-to-text
tags:
- qwen
- qwen3.8
- gguf
- abliterix
- multimodal
- vision-language
- llama.cpp
- mtp
- speculative-decoding
language:
- multilingual
---
Qwen3.8-27B-Abliterated-MTP-GGUF
Unofficial community derivative of Qwen/Qwen3.8-27B, published by GCSA-AiLab.
这是 GCSA-AiLab 发布的 Qwen/Qwen3.8-27B 非官方社区衍生版本。
Dense multimodal derivative with optional MTP.
稠密多模态衍生模型，可选 MTP。
Introduction and usage / 介绍与使用
English | 简体中文 | 繁體中文 | 日本語 | 한국어 | Français | Português
Downloads / 下载
Technical details · English | 技术细节 · 简体中文
Experimental changes to refusal behavior may alter safeguards. Validate outputs and safety measures before use; follow applicable laws and the Apache-2.0 upstream terms.
拒答相关行为的实验性调整可能改变安全保护。使用前请验证输出及安全措施，并遵守适用法律和上游许可证。
Run GCSA-AiLab/Qwen3.8-27B-Abliterated-MTP-GGUF with guIDE
Download guIDE — the AI-native code editor with local LLM inference and 69 built-in tools.
Source: Hugging Face · Compare models