Kimi K3 on a Potato: C99 Runs a 2.78T Model in 8.24GB RAM
Moonshot’s own docs say Kimi K3 needs 16 nodes and about 4.8TB of VRAM. Then I found a 176KB C99 file that runs the same 2.78-trillion-parameter model on a laptop with 8GB of RAM — no GPU, no BLAS, no framework. Slow? 32 seconds per token. And real? The numbers are all in the repo, measured, not promised. I almost scrolled past it, because “runs a frontier model on a potato” is usually marketing. But this one isn’t. FareedKhan-dev’s kimi-k3-in-c hit 607 stars in its first two days — roughly 300 a day, the third-fastest new AI project on GitHub right now — and by the time I checked the API while writing this it was at 633. That pace alone earned a look. The engineering earned the rest of my evening. ...