Audio version — Estimated duration: 1 min 29 sec
Installation Guide
Welcome to the installation guide for llama.cpp. This section will walk you through the process of setting up the environment and building the project for different hardware configurations.
🚀 Getting Started
Before you begin any specific installation method, you must ensure your system meets the necessary requirements.
- Check Prerequisites: Review the Prerequisites Guide to ensure you have Git, CMake, and a C++ compiler installed.
- Choose Your Hardware Path: Select the installation method that matches your computer’s hardware.
🛠️ Installation Methods
Choose the path that best suits your hardware setup:
💻 Standard (CPU Only)
The most compatible method. Works on virtually any modern Windows machine without needing a dedicated graphics card. 👉 Follow the CPU Installation Guide
🟢 NVIDIA GPU Acceleration (CUDA)
Highly recommended if you have an NVIDIA graphics card. Provides significant performance boosts for model inference. 👉 Follow the CUDA Installation Guide
🔴 AMD GPU Acceleration (HIP)
For users with AMD graphics cards. Leverages AMD’s HIP SDK to accelerate computations on your GPU. 👉 Follow the AMD Installation Guide
🔵 Intel GPU Acceleration (SYCL)
For users with Intel Arc, Iris Xe, or UHD Graphics. Uses Intel oneAPI with SYCL. 👉 Follow the Intel Installation Guide
🟣 Vulkan (Universal GPU)
For any GPU that supports Vulkan (NVIDIA, AMD, Intel, and more). A portable fallback that works everywhere. 👉 Follow the Vulkan Installation Guide
🤔 Not Sure Which Backend to Pick?
⏭️ What’s Next?
Once you have successfully completed your installation and verified it works, you can move on to using the tool.
- Running Models: Learn how to use the command line interface in the CLI Usage Guide.
- HTTP Server: Run a ChatGPT-like web interface with llama-server.
- Exploring Models: See what different formats are available in Model Formats.
Last Updated: 2026-07-19