---
title: DeepSeek-V4-Flash-W4AFP8
canonical_url: "https://www.modelscope.ai/models/MooreThreads/DeepSeek-V4-Flash-W4AFP8"
md_url: "https://www.modelscope.ai/models/MooreThreads/DeepSeek-V4-Flash-W4AFP8.md"
repository: MooreThreads/DeepSeek-V4-Flash-W4AFP8
last_updated: 2026-08-05
license: mit
model_type:
  - deepseek_v4
architectures:
  - DeepseekV4ForCausalLM
base_model_relation: quantized
library_name:
  - safetensors
  - pytorch
frameworks:
  - Pytorch
downloads: 104
stars: 0
---

# DeepSeek-V4-Flash-W4AFP8

> DeepSeek-V4-Flash-W4AFP8 - An open-source model by MooreThreads on ModelScope. DeepSeek-V4-Flash-W4AFP8

MooreThreads/DeepSeek-V4-Flash-W4AFP8 is a machine learning model on ModelScope. licensed under mit.

- **Repository**: MooreThreads/DeepSeek-V4-Flash-W4AFP8
- **License**: mit
- **Downloads**: 104
- **Stars**: 0
- **Last updated**: 2026-08-05

Source: https://www.modelscope.ai/models/MooreThreads/DeepSeek-V4-Flash-W4AFP8

---

# DeepSeek-V4-Flash-W4AFP8

## 模型介绍

本模型为 DeepSeek-V4-Flash MoE 量化版本。

## 量化方式

- 原始权重从 FP4 转换为 INT4
- scale 的 group shape 从 32 调整为 128