---
title: DeepSeek-V3.2-Speciale-4bit
canonical_url: "https://www.modelscope.cn/models/mlx-community/DeepSeek-V3.2-Speciale-4bit"
md_url: "https://www.modelscope.cn/models/mlx-community/DeepSeek-V3.2-Speciale-4bit.md"
repository: mlx-community/DeepSeek-V3.2-Speciale-4bit
last_updated: 2025-12-03
license: mit
pipeline_tag: text-generation
tasks:
  - text-generation
model_type:
  - deepseek_v32
architectures:
  - DeepseekV32ForCausalLM
base_model:
  - deepseek-ai/DeepSeek-V3.2-Speciale
base_model_relation: finetune
parameters: 102.7B
tensor_type:
  - BF16
  - U32
  - F32
library_name:
  - mlx
  - safetensors
  - pytorch
frameworks:
  - pytorch
downloads: 143
stars: 1
tags:
  - mlx
---

# DeepSeek-V3.2-Speciale-4bit

> DeepSeek-V3.2-Speciale-4bit - mlx-community 在 ModelScope 开源的模型。mlx-community/DeepSeek-V3.2-Speciale-4bit

mlx-community/DeepSeek-V3.2-Speciale-4bit 是 ModelScope 魔搭社区上的 102.7B 参数text-generation模型，采用 mit 许可，基于 deepseek-ai/DeepSeek-V3.2-Speciale 构建。

- **Repository**: mlx-community/DeepSeek-V3.2-Speciale-4bit
- **License**: mit
- **Tasks**: text-generation
- **Parameters**: 102.7B
- **Base model**: deepseek-ai/DeepSeek-V3.2-Speciale
- **Tags**: mlx
- **Downloads**: 143
- **Stars**: 1
- **Last updated**: 2025-12-03

Source: https://www.modelscope.cn/models/mlx-community/DeepSeek-V3.2-Speciale-4bit

---

# mlx-community/DeepSeek-V3.2-Speciale-4bit

This model [mlx-community/DeepSeek-V3.2-Speciale-4bit](https://huggingface.co/mlx-community/DeepSeek-V3.2-Speciale-4bit) was
converted to MLX format from [deepseek-ai/DeepSeek-V3.2-Speciale](https://huggingface.co/deepseek-ai/DeepSeek-V3.2-Speciale)
using mlx-lm version **0.28.4**.

## Use with mlx

```bash
pip install mlx-lm
```

```python
from mlx_lm import load, generate

model, tokenizer = load("mlx-community/DeepSeek-V3.2-Speciale-4bit")

prompt = "hello"

if tokenizer.chat_template is not None:
    messages = [{"role": "user", "content": prompt}]
    prompt = tokenizer.apply_chat_template(
        messages, add_generation_prompt=True
    )

response = generate(model, tokenizer, prompt=prompt, verbose=True)
```