---
title: "Discussion on \"Two DGX Sparks, One Model: DeepSeek-V4-Flash with a 1M-Token Context\" | Hashnode"
description: "Discussion on \"Two DGX Sparks, One Model: DeepSeek-V4-Flash with a 1M-Token Context\". Two NVIDIA DGX Sparks, one DeepSeek-V4-Flash model, and a full 1M-token context. The story behind the repo, the 30 MiB near-miss, and what I learned."
url: "https://hashnode.com/posts/two-dgx-sparks-one-model-deepseek-v4-flash-with-a-1m-token-context/6aa97ecc9e408ed2c912ffef"
---

[Feed](https://hashnode.com/feed) /

Discussion

# [Two DGX Sparks, One Model: DeepSeek-V4-Flash with a 1M-Token Context](https://dashaun.hashnode.dev/two-dgx-sparks-one-model-deepseek-v4-flash-with-a-1m-token-context?utm_source=hashnode&utm_medium=feed)

Two NVIDIA DGX Sparks, one DeepSeek-V4-Flash model, and a full 1M-token context. The story behind the repo, the 30 MiB near-miss, and what I learned.

Sep 15 · 10 min read · dashaun.hashnode.dev  
Read the article

-   [#dgx-spark](https://hashnode.com/tag/dgx-spark)
-   [#llm-inference](https://hashnode.com/tag/llm-inference)
-   [#homelab](https://hashnode.com/tag/homelab)
-   [#deepseek](https://hashnode.com/tag/deepseek)
-   [#ray](https://hashnode.com/tag/ray)
-   [#vllm](https://hashnode.com/tag/vllm)

## Written by

[DaShaun](https://hashnode.com/@dashaun)

@dashaun

  
[View profile](https://hashnode.com/@dashaun)

Deliberately practicing to build better software, faster.

A husband, father of four, volunteer and struggling athlete. Deliberately practicing to build better software, faster. Doesn't have anything figured out. Trying to get a little better every day.

## Responses

Markdown supported · ⌘↵ to post

No responses yet.
