---
license: mit
tags:
  - gguf
  - llama-cpp
  - windows
  - runtime
---

# llama.cpp b9637 Windows runtime mirror

This repository mirrors an immutable subset of the official
[`ggml-org/llama.cpp` b9637 release](https://github.com/ggml-org/llama.cpp/releases/tag/b9637)
for resilient application installation on Windows x64.

Included archives:

- Windows x64 CPU
- Windows x64 Vulkan
- Windows x64 CUDA 12.4 executable/runtime files built by llama.cpp
- Windows x64 CUDA 13.3 executable/runtime files built by llama.cpp

Every file is unmodified. Verify it against `checksums.sha256` before use.

NVIDIA CUDA runtime and cuBLAS redistributable archives are deliberately not
republished here. Applications that select a CUDA build must obtain those
components from the official upstream release under the NVIDIA CUDA EULA, or
fall back to the Vulkan build.

This mirror is not affiliated with or endorsed by the llama.cpp project or
NVIDIA.