installama.sh

The fastest way to install llama.cpp on Linux and macOS.

installama.sh is a simple shell script that downloads and sets up a prebuilt llama-server binary for your system. It automatically detects your OS, architecture, and GPU capabilities, so you can start using llama.cpp in seconds.

Features

• Automatic detection of CPU architecture (x86_64 / aarch64) and OS (Linux / macOS).
• Support for GPU acceleration:
  • CUDA: 50 61 70 75 80 86 89.
  • ROCm: gfx803 gfx900 gfx906 gfx908 gfx90a gfx942 gfx1010 gfx1011 gfx1030 gfx1032 gfx1100 gfx1101 gfx1102 gfx1200 gfx1201 gfx1151.
  • Metal: M1 M2 M3 M4.
• Fallback to CPU optimized builds if the GPU is not available.
• Lightweight and Fast!

Usage

Install llama-server in one easy step:

curl angt.github.io/installama.sh | sh

Then run the server, for example, with the new awesome WebGUI:

~/.installama/llama-server -hf ggml-org/gpt-oss-20b-GGUF --jinja

And open your favorite browser to http://127.0.0.1:8080/.

You can also run everything in a single command:

curl angt.github.io/installama.sh | MODEL=ggml-org/gpt-oss-20b-GGUF sh

In some scenarios, you may want to skip the CUDA backend. You can do this with the following command:

curl angt.github.io/installama.sh | SKIP_CUDA=1 sh

Skipping ROCm is also possible by setting SKIP_ROCM=1.

---

⚠️ This is still a PoC. If it doesn't work on your system, please create an issue.