kimi-k3-in-c
Summary
FareedKhan-dev's repository documents building a 2.78-trillion-parameter Kimi K3 model in C99 to run inference on a single CPU with about 8.24 GB of RAM. The project emphasizes a portable, zero-dependency approach (no BLAS, no GPU) and provides structure for benchmarks, docs, and tooling around CPU-based inference for large language models.