README.md

---
license: mit
base_model:
- zai-org/AutoGLM-Phone-9B
tags:
- multimodal
- vision-language
- phone-automation
- gguf
- quantized
- agent
language:
- zh
pipeline_tag: image-text-to-text
---

# AutoGLM-Phone-9B GGUF Quantized Model Collection/AutoGLM-Phone-9B GGUF 量化模型集合

Congratulations! This is the most complete and fully usable collection of AutoGLM-Phone-9B model GGUF quantized versions you can find.🎉🎉🎉   
恭喜你！这是你能找到最完整，并且绝对可用的 AutoGLM-Phone-9B 模型 GGUF 量化版本集合。🎉🎉🎉

## Model Introduction/模型简介

Phone Agent is a mobile intelligent assistant framework built on AutoGLM, 
capable of understanding smartphone screens through multimodal perception and executing automated operations to complete tasks.   
AutoGLM-Phone-9B 是基于 GLM-4V-9B 的多模态视觉语言模型，专门针对手机自动化场景进行了优化。该模型能够理解手机屏幕截图并生成相应的操作指令。

⚠️Please note! This is a multimodal vision language model, 
so in addition to the model itself, you also need the mmproj file. Please be sure to download this file for use!   
⚠️请注意！这是多模态视觉语言模型，所以除了模型本身，你还需要mmproj文件，请务必下载这个文件一起使用！

## Available quantization versions/可用的量化版本

| Quantization Type | Size | Memory Requirement | Notes | Download Link |
|---------|---------|---------|----------|----------|
| **Q2_K** | 3.73 GB | ~4 GB | Not recommended 不推荐 | [Download](AutoGLM-Phone-9B-Q2_K.gguf) |
| **Q3_K_S** | 4.28 GB | ~5 GB | Not recommended 不推荐 | [Download](AutoGLM-Phone-9B-Q3_K_S.gguf) |
| **Q3_K_M** | 4.63 GB | ~5 GB | Lower quality 质量较低 | [Download](AutoGLM-Phone-9B-Q3_K_M.gguf) |
| **Q3_K_L** | 4.84 GB | ~6 GB | Lower quality 质量较低 | [Download](AutoGLM-Phone-9B-Q3_K_L.gguf) |
| **Q4_0** | 5.08 GB | ~6 GB | Minimum available 最低可用 | [Download](AutoGLM-Phone-9B-Q4_0.gguf) |
| **Q4_1** | 5.60 GB | ~6 GB | 	Fast, recommended 快速，推荐 | [Download](AutoGLM-Phone-9B-Q4_1.gguf) |
| **Q4_K_S** | 5.36 GB | ~6 GB | Fast, recommended 快速，推荐 | [Download](AutoGLM-Phone-9B-Q4_K_S.gguf) |
| **Q4_K_M** | 5.74 GB | ~7 GB | ⭐Most Recommended, balanced 最推荐，平衡⭐ | [Download](AutoGLM-Phone-9B-Q4_K_M.gguf) |
| **Q5_0** | 6.11 GB | ~7 GB | Not recommended 不推荐 | [Download](AutoGLM-Phone-9B-Q5_0.gguf) |
| **Q5_1** | 6.62 GB | ~8 GB | Not recommended 不推荐 | [Download](AutoGLM-Phone-9B-Q5_1.gguf) |
| **Q5_K_S** | 6.24 GB | ~7 GB | Good quality 质量不错 | [Download](AutoGLM-Phone-9B-Q5_K_S.gguf) |
| **Q5_K_M** | 6.57 GB | ~8 GB | Good quality 质量不错 | [Download](AutoGLM-Phone-9B-Q5_K_M.gguf) |
| **Q6_K** | 7.70 GB | ~9 GB | Very good quality  质量非常好 | [Download](AutoGLM-Phone-9B-Q6_K.gguf) |
| **Q8_0** | 9.31 GB | ~11 GB | ⭐Fast, best quality 快速，质量最好⭐ | [Download](AutoGLM-Phone-9B-Q8_0.gguf) |
| **F16** | 17.52 GB | ~20 GB | 16 bpw, overkill  16 bpw，过量 | [Download](AutoGLM-Phone-9B-F16.gguf) |

## Quick Start/快速开始

### Using llama.cpp/使用 llama.cpp

```bash
# Download the model and visual projector
# 下载模型和视觉投影器
wget https://huggingface.co/BlcaCola/AutoGLM-Phone-9B-GGUF.gguf/resolve/main/AutoGLM-Phone-9B-Q8_0.gguf
wget https://huggingface.co/BlcaCola/AutoGLM-Phone-9B-GGUF.gguf/resolve/main/AutoGLM-Phone-9B-mmproj.gguf

# Start Server
# 启动服务器
./llama-server -m AutoGLM-Phone-9B-Q8_0.gguf --mmproj AutoGLM-Phone-9B-mmproj.gguf --host 0.0.0.0 --port 8080
```


## Performance Comparison/性能对比

Here is a chart by ikawrakow comparing the performance levels of partially quantized models (below Q5):   
这里有一张 ikawrakow 的图表，比较了部分量化的性能水平（低于Q5）：

![image.png](https://www.nethype.de/huggingface_embed/quantpplgraph.png)

## Related Resources/相关资源

- **原项目/Original project**: [Open-AutoGLM](https://github.com/zai-org/Open-AutoGLM)
- **llama.cpp**: [GitHub](https://github.com/ggerganov/llama.cpp)

## License Agreement/使用许可

This model is licensed under the MIT License. Please refer to the license terms of the original model.   
本模型遵循 MIT 许可证。请查看原始模型的许可证条款。


---