I pushed Kimi K3 onto one CPU with 8 GB of RAM

I deployed K3 on 32 H100s at work a couple of weeks ago and then got annoyed that there was no way to poke at it on my own machine. So I wrote an inference engine for it in C99. Nothing clever going on. 93% of that 1.56 TB checkpoint is routed experts, and only 16 of 896 fire per token, so the exper

中文全文翻译正在补齐，可先查看下方英文原文。

I deployed K3 on 32 H100s at work a couple of weeks ago and then got annoyed that there was no way to poke at it on my own machine. So I wrote an inference engine for it in C99. Nothing clever going on. 93% of that 1.56 TB checkpoint is routed experts, and only 16 of 896 fire per token, so the exper