#c_lang #avx2 #c99 #cpu_inference #deep_learning #from_scratch #inference_engine #kimi_k3 #linear_attention #llm #llm_inference #machine_learning #memory_efficient #mixture_of_experts #moe #mxfp4 #quantization #simd #systems_programming #transformer #zero_dependencies
kimi-k3-in-c runs the Kimi K3 language model on a CPU without a GPU or machine learning framework. It supports text generation and multi-turn text chat on Linux, macOS, and Windows. By loading model data from disk as needed, it can run the full model with limited memory, making it useful on machines that cannot hold the model in RAM. It requires roughly 1.7 TB of free storage, and generation can be slow, especially with less memory or slower disks.
https://github.com/FareedKhan-dev/kimi-k3-in-c
kimi-k3-in-c runs the Kimi K3 language model on a CPU without a GPU or machine learning framework. It supports text generation and multi-turn text chat on Linux, macOS, and Windows. By loading model data from disk as needed, it can run the full model with limited memory, making it useful on machines that cannot hold the model in RAM. It requires roughly 1.7 TB of free storage, and generation can be slow, especially with less memory or slower disks.
https://github.com/FareedKhan-dev/kimi-k3-in-c
GitHub
GitHub - FareedKhan-dev/kimi-k3-in-c: A 2.78-trillion-parameter Kimi K3 running inference on a single CPU in 8.24 GB of RAM. Portable…
A 2.78-trillion-parameter Kimi K3 running inference on a single CPU in 8.24 GB of RAM. Portable C99: no BLAS, no framework, no GPU. - FareedKhan-dev/kimi-k3-in-c