FareedKhan-dev/kimi-k3-in-c
kimi-k3-in-c runs a 2.78-trillion-parameter model entirely on a single CPU. It is built for developers who want to execute massive inference without GPUs or external frameworks.
10.1k · C
The codebase is written in portable C99 and fits into 8.24 GB of RAM without requiring BLAS. If your infrastructure already relies on heavy GPU clusters or distributed multi-node setups for standard production traffic, this lightweight CPU-only runner will bring no practical benefits because your hardware requirements are already met elsewhere.