Audio version — Estimated duration: 1 min 54 sec

Troubleshooting

Running into problems? This section covers the most common issues encountered when building and running llama.cpp.

📄 Pages in This Section

Common Issues & Solutions — Fix build errors, runtime crashes, and performance problems.

Problem	Likely Fix
`cmake` not found	Install CMake, add to PATH, restart terminal
`git` not found	Install Git, add to PATH, restart terminal
CUDA not found	Install CUDA Toolkit or use `-DGGML_CUDA=OFF`
`failed to allocate`	Reduce `-ngl` or use a smaller quantization
Model gives gibberish	Use the correct prompt template for your model
Very slow generation	Use `-ngl` (GPU offload) or reduce context size `-c`

Last Updated: 2026-07-19