# [260803] A 13-Year-Old Server Running a 26B Model: AI Is Redefining Hardware Value

> Source: <https://x-cmd.com/blog/260803/>
> Published: 2026-08-03 00:00:00+00:00

**TLDR**

- Developer Ryan Findley ran Google's Gemma 4 26B MoE model on a 2013 HP StoreVirtual server (dual Xeon E5 + DDR3, no GPU). After quantization and inference tuning, it hits about 5.2 tokens/sec, and the whole used server costs under $300. It can't train large models, but for a personal assistant, document summarization, coding help, or a local knowledge base, it's already "good enough." The same pattern has played out before: Linux turned commodity x86 boxes into enterprise servers, and OpenWrt gave discontinued routers a second life. AI is rewriting the rule we used to judge machines by — hardware sets the ceiling; software decides how much of that ceiling a machine actually gets to use.

[Read more](https://x-cmd.com/blog/260803/)