---
title: DeepSeek-V4-Flash
canonical_url: "https://www.modelscope.cn/models/mlx-community/DeepSeek-V4-Flash"
md_url: "https://www.modelscope.cn/models/mlx-community/DeepSeek-V4-Flash.md"
repository: mlx-community/DeepSeek-V4-Flash
last_updated: 2026-06-07
license: mit
pipeline_tag: text-generation
tasks:
  - text-generation
model_type:
  - deepseek_v4
architectures:
  - DeepseekV4ForCausalLM
base_model:
  - deepseek-ai/DeepSeek-V4-Flash
base_model_relation: quantized
parameters: 45.4B
tensor_type:
  - BF16
  - F32
  - U8
  - I64
  - U32
library_name:
  - mlx
  - safetensors
  - pytorch
frameworks:
  - pytorch
language:
  - en
downloads: 181
stars: 0
tags:
  - mlx
---

# DeepSeek-V4-Flash

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

mlx-community/DeepSeek-V4-Flash 是 ModelScope 魔搭社区上的 45.4B 参数text-generation模型，采用 mit 许可，基于 deepseek-ai/DeepSeek-V4-Flash 构建。

- **Repository**: mlx-community/DeepSeek-V4-Flash
- **License**: mit
- **Tasks**: text-generation
- **Parameters**: 45.4B
- **Base model**: deepseek-ai/DeepSeek-V4-Flash
- **Tags**: mlx
- **Downloads**: 181
- **Stars**: 0
- **Last updated**: 2026-06-07

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

---

# mlx-community/DeepSeek-V4-Flash

This 8-bit model [mlx-community/DeepSeek-V4-Flash](https://huggingface.co/mlx-community/DeepSeek-V4-Flash) was converted to MLX format from [deepseek-ai/DeepSeek-V4-Flash](https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash).

DeepSeek-V4-Flash is a Mixture-of-Experts model with 284B total parameters, 13B activated parameters, and a context length of 1M tokens. Refer to the [original model card](https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash) for more details on the model.

## Use with mlx

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

```python
from mlx_lm import load, generate

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

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)
```