llama-cpp-python — Free-Tier Friendly Wheel

This Space provides a prebuilt llama-cpp-python wheel designed to work reliably on Hugging Face Free tier Spaces.

No compilation. No system packages. No build failures.

If your Space crashes during pip install llama-cpp-python, this wheel is the fix.

Optimized for Hugging Face Free Tier

Hugging Face Free tier Spaces are:

CPU-only
Limited in memory
Not suitable for native compilation

This wheel is built ahead of time so it can be installed instantly without triggering CMake, compilers, or BLAS detection.

What this wheel gives you

✅ Works on HF Free tier CPU Spaces
✅ Linux (ubuntu-22.04 compatible)
✅ Python 3.10
✅ OpenBLAS enabled (GGML_BLAS=ON)
✅ No system dependencies required
✅ No build step during Space startup
✅ Fast, reliable pip install

How to use in a Space (Free tier)

Download the wheel from the GitHub repository
Upload it to your Space
Install it in your Space startup:

pip install llama_cpp_python-*.whl>

That’s it — your Space will start without build errors.

Build details

This wheel was built using: abetlen/llama-cpp-python (recursive submodules) OpenBLAS (GGML_VENDOR=OpenBLAS) scikit-build-core ninja python -m build --wheel --no-isolation

Build environment:

OS: Ubuntu 22.04 Python: 3.10

Why not build from source on HF?

On Free tier Spaces, building from source often fails due to: Missing compilers Missing BLAS libraries Memory limits Build timeouts This prebuilt wheel avoids all of those issues.

Notes

CPU-only (no CUDA) Intended for inference workloads Not an official upstream release

Credits

All credit goes to the maintainers of llama-cpp-python and llama.cpp. This Space exists solely to make Free tier usage painless.