---
title: DeepSeek-V4-Flash-2bit-DQ
canonical_url: "https://www.modelscope.cn/models/mlx-community/DeepSeek-V4-Flash-2bit-DQ"
md_url: "https://www.modelscope.cn/models/mlx-community/DeepSeek-V4-Flash-2bit-DQ.md"
repository: mlx-community/DeepSeek-V4-Flash-2bit-DQ
last_updated: 2026-04-28
pipeline_tag: text-generation
tasks:
  - text-generation
model_type:
  - deepseek_v4
architectures:
  - DeepseekV4ForCausalLM
parameters: 30.0B
tensor_type:
  - F32
  - BF16
  - U32
  - I64
library_name:
  - mlx
  - safetensors
  - pytorch
frameworks:
  - pytorch
language:
  - en
downloads: 247
stars: 0
tags:
  - mlx
---

# DeepSeek-V4-Flash-2bit-DQ

> DeepSeek-V4-Flash-2bit-DQ - mlx-community 在 ModelScope 开源的模型。mlx-community/DeepSeek-V4-Flash-2bit-DQ

- **Repository**: mlx-community/DeepSeek-V4-Flash-2bit-DQ
- **Tasks**: text-generation
- **Parameters**: 30.0B
- **Tags**: mlx
- **Downloads**: 247
- **Stars**: 0
- **Last updated**: 2026-04-28

Source: https://www.modelscope.cn/models/mlx-community/DeepSeek-V4-Flash-2bit-DQ

---

# mlx-community/DeepSeek-V4-Flash-2bit-DQ

Made possible by [Lambda.ai](https://huggingface.co/lambda) ❤️

DeepSeek-V4-Flash-2bit-DQ uses a dynamic mixed-precision quantization policy. Most routed MoE expert weights are packed to 2-bit, while sensitive layers and projections remain in higher-quality 4-bit, 6-bit or 8-bit quantization. This keeps memory use much lower than the baseline 4-bit checkpoint.

## Use with mlx

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

```python
from mlx_lm import load, generate

model, tokenizer = load("mlx-community/DeepSeek-V4-Flash-2bit-DQ")

prompt = "hello"

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

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