# Kimi K3 in C

> A Kimi K3 inference engine in C99, dependency-free, on CPU

- Category: Agents & dev (`agents`)
- Tags: `open-source`, `free`
- Official: https://github.com/FareedKhan-dev/kimi-k3-in-c
- Wunderlust: https://wunderlust.xyz/r/kimi-k3-in-c
- Image: https://wunderlust.xyz/media/kimi-k3-in-c-poster.jpg
- Video preview: https://wunderlust.xyz/media/kimi-k3-in-c.webm
- Added: 2026-11-05

## What it is

A pure C99 open source implementation of inference for the Kimi K3 model (2.78 trillion parameters) on a single CPU, with no BLAS, framework or GPU.

## What it’s for

A reference for studying large-model inference on CPU with MXFP4 quantization and AVX2 optimizations; not a general-purpose tool.

## Benefits

- Portable C99 code with CMake and a Makefile, no dependencies
- Mixture-of-experts architecture with linear attention and MXFP4 quantization

## Caveats

- Apache-2.0 license
- The repository cites about 8 GB of RAM in the test and a checkpoint of roughly 1.56 TB on disk; that doesn't mean it runs on any machine with 8 GB

## Similar on Wunderlust

- [Qwen3.8-Max](https://wunderlust.xyz/r/qwen3-8-max.md): Repository for Qwen3.8-Max, Alibaba Cloud's MoE model
- [Ollama](https://wunderlust.xyz/r/ollama.md): Runtime for running language models on your own machine
- [MLX](https://wunderlust.xyz/r/mlx.md): Stack for running local models on Macs
- [Unsloth Studio](https://wunderlust.xyz/r/unsloth-studio.md): App for running, chatting with and training AI models locally
- [GitDiagram](https://wunderlust.xyz/r/gitdiagram.md): Generates diagrams from code repositories
- [Personal AI Infrastructure (PAI)](https://wunderlust.xyz/r/personal-ai-infrastructure.md): Personal AI infrastructure approach with memory, hooks and autonomy