How to beginner · 3 min read

How to compile llama.cpp

Q: How to compile llama.cpp

To compile llama.cpp, clone the repository, install dependencies like cmake and a C++ compiler, then run cmake . followed by make in the project directory. This builds the executable for use with ollama or local inference.

Quick answer

To compile llama.cpp, clone the repository, install dependencies like cmake and a C++ compiler, then run cmake . followed by make in the project directory. This builds the executable for use with ollama or local inference.

PREREQUISITES

C++17 compatible compiler (gcc, clang, or MSVC)
CMake 3.15+ installed
Git installed
Make (Linux/macOS) or build tools for Windows
Basic terminal/command line knowledge

Setup

First, ensure you have git, cmake, and a C++17 compiler installed. On macOS, install Xcode Command Line Tools and cmake via Homebrew. On Linux, use your package manager to install build-essential and cmake. On Windows, install Visual Studio with C++ workload and CMake.

bash

git --version
cmake --version
clang --version  # or gcc --version

output

git version 2.40.1
cmake version 3.26.4
Apple clang version 14.0.0 (clang-1400.0.29.202)

Step by step

Clone the llama.cpp repository and build it using cmake and make:

bash

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake .
make

output

Scanning dependencies of target llama
[ 50%] Building CXX object CMakeFiles/llama.dir/llama.cpp.o
[100%] Linking CXX executable llama
[100%] Built target llama

Common variations

For Windows, use Visual Studio to open the generated llama.sln file and build the project. To build with specific options, pass flags to cmake, e.g., cmake -DLLAMA_CUBLAS=ON . for CUDA support. You can also build with ninja if preferred.

bash

cmake -G Ninja .
ninja

output

[100%] Built target llama

Troubleshooting

If cmake fails, verify your compiler supports C++17 and your environment variables are set correctly.
On Windows, ensure the Visual Studio Developer Command Prompt is used.
If make is not found, install it or use an alternative build system like Ninja.

✅

Key Takeaways

Use CMake and a C++17 compiler to build llama.cpp from source.
On Windows, prefer Visual Studio or Ninja for building.
Check environment and dependencies if build errors occur.

Verified 2026-04

Verify ↗