---
title: DeepSeek-V4-Flash-0731-EW-MFQ
canonical_url: "https://www.modelscope.cn/models/Tylogi/DeepSeek-V4-Flash-0731-EW-MFQ"
md_url: "https://www.modelscope.cn/models/Tylogi/DeepSeek-V4-Flash-0731-EW-MFQ.md"
repository: Tylogi/DeepSeek-V4-Flash-0731-EW-MFQ
last_updated: 2026-08-07
license: mit
pipeline_tag: text-generation
tasks:
  - text-generation
base_model_relation: quantized
downloads: 0
stars: 0
tags:
  - mfq
  - quantization
  - deepseek_v4
  - deepseek
---

# DeepSeek-V4-Flash-0731-EW-MFQ

> DeepSeek-V4-Flash-0731-EW-MFQ - Tylogi 在 ModelScope 开源的模型。Run DeepSeek-V4-Flash-0731 with TyloQuant MFQ

Tylogi/DeepSeek-V4-Flash-0731-EW-MFQ 是 ModelScope 魔搭社区上的text-generation模型，采用 mit 许可。

- **Repository**: Tylogi/DeepSeek-V4-Flash-0731-EW-MFQ
- **License**: mit
- **Tasks**: text-generation
- **Tags**: mfq, quantization, deepseek_v4, deepseek
- **Downloads**: 0
- **Stars**: 0
- **Last updated**: 2026-08-07

Source: https://www.modelscope.cn/models/Tylogi/DeepSeek-V4-Flash-0731-EW-MFQ

---

## Run DeepSeek-V4-Flash-0731 with TyloQuant MFQ

<p style="margin-top: 0; margin-bottom: 8px;">
  <em>TyloQuant MFQ provides neuron-anchored mixed-format quantization and high-fidelity inference for large language models.</em>
</p>

<div style="display: flex; gap: 8px; align-items: center; flex-wrap: wrap;">
  <a href="https://github.com/Tylogi/TyloQuant">
    <img src="https://img.shields.io/badge/GitHub-TyloQuant-172033?logo=github&logoColor=white" alt="TyloQuant GitHub">
  </a>
  <a href="https://github.com/Tylogi/TyloQuant#readme">
    <img src="https://img.shields.io/badge/Documentation-English-2477d4" alt="English documentation">
  </a>
  <a href="https://github.com/Tylogi/TyloQuant/blob/master/README.zh-CN.md">
    <img src="https://img.shields.io/badge/Documentation-中文-2477d4" alt="中文文档">
  </a>
  <a href="https://github.com/Tylogi/TyloQuant/blob/master/docs/deepseek-v4-flash-0731-results.md">
    <img src="https://img.shields.io/badge/Benchmark-Full_results-e07a2e" alt="Full benchmark results">
  </a>
</div>

<div align="center">
  <a href="https://github.com/Tylogi/TyloQuant">
    <img src="https://github.com/Tylogi/TyloQuant/raw/master/docs/figures/tylogi-ai-lab.svg" width="520" alt="Tylogi AI Lab">
  </a>
</div>

<ul style="margin-top: 8px;">
  <li>This repository contains MFQ-quantized weights derived from the official <b>DeepSeek-V4-Flash-0731</b> release.</li>
  <li>MFQ combines NINT, NVQ/NPQ and NEPQ formats with expert-wise precision allocation for MoE models.</li>
  <li>The released 77.519 GiB S tier records 0.313576 Mean KLD and 82.2913% same-top on the complete official-0731 WikiText-2 evaluation.</li>
  <li>See the <a href="https://github.com/Tylogi/TyloQuant">project documentation</a> for runtime, format and deployment instructions.</li>
</ul>

<img src="./assets/deepseek-v4-flash-mfq-vs-ud-kld.png" alt="DeepSeek-V4-Flash-0731 MFQ versus Unsloth Dynamic model size and Mean KLD" width="100%">

<p><sub>Full WikiText-2 evaluation against official DeepSeek-V4-Flash-0731 BF16 logits: ctx=512, 573 chunks and 146,115 scored tokens. Lower Mean KLD is better. <a href="https://github.com/Tylogi/TyloQuant/blob/master/docs/deepseek-v4-flash-0731-results.md">Complete results and protocol</a>.</sub></p>

---

# DeepSeek-V4-Flash-0731

<!-- markdownlint-disable first-line-h1 -->
<!-- markdownlint-disable html -->
<!-- markdownlint-disable no-duplicate-header -->

<hr>
<div align="center" style="line-height: 1;">
  <a href="https://www.deepseek.com/" target="_blank" style="margin: 2px;">
    <img alt="Homepage" src="https://github.com/deepseek-ai/DeepSeek-V2/blob/main/figures/badge.svg?raw=true" style="display: inline-block; vertical-align: middle;"/>
  </a>
  <a href="https://chat.deepseek.com/" target="_blank" style="margin: 2px;">
    <img alt="Chat" src="https://img.shields.io/badge/🤖%20Chat-DeepSeek%20V4-536af5?color=536af5&logoColor=white" style="display: inline-block; vertical-align: middle;"/>
  </a>
</div>
<div align="center" style="line-height: 1;">
  <a href="https://huggingface.co/deepseek-ai" target="_blank" style="margin: 2px;">
    <img alt="Hugging Face" src="https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-DeepSeek%20AI-ffc107?color=ffc107&logoColor=white" style="display: inline-block; vertical-align: middle;"/>
  </a>
  <a href="https://twitter.com/deepseek_ai" target="_blank" style="margin: 2px;">
    <img alt="Twitter Follow" src="https://img.shields.io/badge/Twitter-deepseek_ai-white?logo=x&logoColor=white" style="display: inline-block; vertical-align: middle;"/>
  </a>
</div>
<div align="center" style="line-height: 1;">
  <a href="LICENSE" style="margin: 2px;">
    <img alt="License" src="https://img.shields.io/badge/License-MIT-f5de53?&color=f5de53" style="display: inline-block; vertical-align: middle;"/>
  </a>
</div>

<p align="center">
  <a href="https://arxiv.org/abs/2606.19348"><b>Technical Report</b>👁️</a>
</p>

## Introduction

**DeepSeek-V4-Flash-0731** is the official release of **DeepSeek-V4-Flash**, superseding the preview version, with substantially enhanced agentic capabilities. It has the same model structure as [DeepSeek-V4-Flash-DSpark](https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-DSpark), i.e. it comes with a speculative decoding module attached.

DeepSeek-V4-Flash-0731 outperforms DeepSeek-V4-Pro (Preview) on benchmarks listed below despite its far smaller activated parameter count, and is broadly competitive with the strongest proprietary models available.

<div align="center">

| Benchmark | DeepSeek-V4-Flash-0731 | DeepSeek-V4-Flash (Preview) | DeepSeek-V4-Pro (Preview) | GLM-5.2 | Opus-4.8 |
| :--- | :---: | :---: | :---: | :---: | :---: |
| Terminal Bench 2.1 | 82.7 | 61.8 | 72.1 | 81.0 | 85.0 |
| NL2Repo | 54.2 | 39.4 | 38.5 | 48.9 | 69.7 |
| Cybergym | 76.7 | 38.7 | 52.7 | - | 83.1 |
| DeepSWE | 54.4 | 7.3 | 12.8 | 46.2 | 58.0 |
| Toolathlon-Verified | 70.3 | 49.7 | 55.9 | 59.9 | 76.2 |
| Agents' Last Exam | 25.2 | 15.8 | 16.5 | 23.8 | 25.7 |
| AutomationBench Public | 25.1 | 10.8 | 12.8 | 12.9 | 27.2 |
| DSBench-FullStack † | 68.7 | 37.0 | 41.8 | 61.8 | 71.6 |
| DSBench-Hard † | 59.6 | 25.8 | 31.1 | 54.5 | 71.7 |

</div>

Notes:

1. For the Code Agent tasks among the public benchmarks above, DeepSeek-V4-Flash-0731 is evaluated with the minimal mode of DeepSeek Harness (to be released) as the agent framework, using the `max` reasoning effort level with `temperature = 1.0, top_p = 0.95`.
2. † DSBench-FullStack is an internal full-stack development test set; DSBench-Hard is an internal test set of difficult coding-agent problems.

## License

This repository and the model weights are licensed under the [MIT License](LICENSE).

## Citation

```
@misc{deepseekai2026deepseekv4,
      title={DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence},
      author={DeepSeek-AI},
      year={2026},
}
```