---
format: "aidr-story-markdown/v1"
id: "55858f2b4ca73bdea18bb38cb3df7729963affc8f5ea52147479775223be0c4a"
canonical_url: "https://aidr.today/55858f2b?lang=en"
title: "Llama-modchip: 284B DeepSeek on 64GB RAM, no GPU"
lang: "en"
requested_lang: "en"
available_langs: ["en","vi"]
translation_fallback: null
fallback_fields: []
published_at: "2026-11-08T10:22:04.000Z"
category: "Open Source"
topics: ["deepseek","llm","open-source"]
source_urls: ["https://github.com/synchopate/llama-modchip","https://news.ycombinator.com/item?id=50411961"]
summary: "Llama-modchip allows running 284B DeepSeek-V4-Flash and 30B MoEs on a 64GB CPU with no GPUs. The setup supports a 256K context window."
---

# Llama\-modchip: 284B DeepSeek on 64GB RAM, no GPU

> [Open the canonical story](<https://aidr.today/55858f2b?lang=en>)

**Published:** 2026-11-08T10:22:04.000Z
**Category:** Open Source
**Topics:** deepseek, llm, open\-source

## Summary

Llama\-modchip allows running 284B DeepSeek\-V4\-Flash and 30B MoEs on a 64GB CPU with no GPUs\. The setup supports a 256K context window\.

## Sources

- [Story source](<https://github.com/synchopate/llama-modchip>)
- [Discussion](<https://news.ycombinator.com/item?id=50411961>)