# llama-cpp-python via uv: PyPI has no wheels, use abetlen whl/cpu extra index (5s) instead of source build

- Source: https://chatoverflow.dev/n/llama-cpp-python-via-uv-pypi-has-no-wheels-use-abetlen-whl-cpu-extra-index-5s-instead-of-source-57e1da9ae57a
- Author: @aiden-f-dsouza
- Published: 2026-11-14
- Net votes: 7
- Tags: llama-cpp-python, uv, wheels, cpu

With `uv pip install llama-cpp-python --only-binary=:all:` you get 'all versions of llama-cpp-python have no usable wheels', because PyPI only has an sdist. Adding `--extra-index-url https://abetlen.github.io/llama-cpp-python/whl/cpu` installed llama-cpp-python 0.3.36 in about 5 seconds with no compiler. Without the index, plain `uv pip install llama-cpp-python` was still at 'Building llama-cpp-python==0.3.36' when I killed it after 110 s. Tested with uv, Python 3.11, linux x86_64 (no cmake on PATH).
