---
title: DeepSeek-V4-Flash-0731-FP8-mt
canonical_url: "https://www.modelscope.cn/models/MooreThreads/DeepSeek-V4-Flash-0731-FP8-mt"
md_url: "https://www.modelscope.cn/models/MooreThreads/DeepSeek-V4-Flash-0731-FP8-mt.md"
repository: MooreThreads/DeepSeek-V4-Flash-0731-FP8-mt
last_updated: 2026-08-10
license: mit
pipeline_tag: text-generation
tasks:
  - text-generation
model_type:
  - deepseek_v4
architectures:
  - DeepseekV4ForCausalLM
base_model:
  - deepseek-ai/DeepSeek-V4-Flash-0731
base_model_relation: quantized
parameters: 304.2B
tensor_type:
  - F8_E4M3
  - F32
  - BF16
  - I64
library_name:
  - safetensors
downloads: 385
stars: 0
tags:
  - deepseek-v4
  - fp8
  - quantized
---

# DeepSeek-V4-Flash-0731-FP8-mt

> DeepSeek-V4-Flash-0731-FP8-mt - MooreThreads 在 ModelScope 开源的模型。DeepSeek-V4-Flash-0731-FP8-mt

MooreThreads/DeepSeek-V4-Flash-0731-FP8-mt 是 ModelScope 魔搭社区上的 304.2B 参数text-generation模型，采用 mit 许可，基于 deepseek-ai/DeepSeek-V4-Flash-0731 构建。

- **Repository**: MooreThreads/DeepSeek-V4-Flash-0731-FP8-mt
- **License**: mit
- **Tasks**: text-generation
- **Parameters**: 304.2B
- **Base model**: deepseek-ai/DeepSeek-V4-Flash-0731
- **Tags**: deepseek-v4, fp8, quantized
- **Downloads**: 385
- **Stars**: 0
- **Last updated**: 2026-08-10

Source: https://www.modelscope.cn/models/MooreThreads/DeepSeek-V4-Flash-0731-FP8-mt

---

# DeepSeek-V4-Flash-0731-FP8-mt

FP8 re-packaging of [`deepseek-ai/DeepSeek-V4-Flash-0731`](https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-0731).
Model architecture, tokenizer, chat template, and reference `encoding/` are
unchanged from the base repo. No fine-tuning, no retraining — weights only.