---
title: DeepSeek-V4-Flash-FP8-Q4
canonical_url: "https://www.modelscope.cn/models/huangyuyang/DeepSeek-V4-Flash-FP8-Q4"
md_url: "https://www.modelscope.cn/models/huangyuyang/DeepSeek-V4-Flash-FP8-Q4.md"
repository: huangyuyang/DeepSeek-V4-Flash-FP8-Q4
last_updated: 2026-05-01
license: mit
model_type:
  - deepseek_v4
architectures:
  - DeepseekV4ForCausalLM
base_model:
  - deepseek-ai/DeepSeek-V4-Flash
base_model_relation: quantized
parameters: 290.9B
tensor_type:
  - F32
  - F16
  - fastllm
library_name:
  - safetensors
  - pytorch
frameworks:
  - pytorch
downloads: 2429
stars: 8
tags:
  - deepseek-v4
  - "fp8 + Q4"
  - quantized
---

# DeepSeek-V4-Flash-FP8-Q4

> DeepSeek-V4-Flash-FP8-Q4 - huangyuyang 在 ModelScope 开源的模型。DeepSeek-V4-Flash-FP8-Q4

huangyuyang/DeepSeek-V4-Flash-FP8-Q4 是 ModelScope 魔搭社区上的 290.9B 参数机器学习模型，采用 mit 许可，基于 deepseek-ai/DeepSeek-V4-Flash 构建。

- **Repository**: huangyuyang/DeepSeek-V4-Flash-FP8-Q4
- **License**: mit
- **Parameters**: 290.9B
- **Base model**: deepseek-ai/DeepSeek-V4-Flash
- **Tags**: deepseek-v4, fp8 + Q4, quantized
- **Downloads**: 2429
- **Stars**: 8
- **Last updated**: 2026-05-01

Source: https://www.modelscope.cn/models/huangyuyang/DeepSeek-V4-Flash-FP8-Q4

---

# DeepSeek-V4-Flash-FP8-Q4

FP8 and Q4(for moe) re-packaging of [`deepseek-ai/DeepSeek-V4-Flash`](https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash).
Model architecture, tokenizer, chat template, and reference `encoding/` are
unchanged from the base repo. No fine-tuning, no retraining — weights only.

# Deployment

Use www.github.com/ztxz16/fastllm

## License

MIT — see `LICENSE`. Copyright © DeepSeek.