When you pip install a package with compiled code, the wheel you get is built for CPU features from 2009. Want newer optimizations like AVX2? Your installer has no way to ask for them. GPU support? You're on your own configuring special index URLs. The result is fat binaries, nearly gigabyte-sized wheels, and install pages that read like puzzle books. A coalition from NVIDIA, Astral, and QuanSight has been working on Wheel Next: A set of PEPs that let packages declare what hardware they need and let installers like uv pick the right build automatically. Just uv pip install torch and it works. I sit down with Jonathan Dekhtiar from NVIDIA, Ralf Gommers from Quansight and the NumPy and SciPy teams, and Charlie Marsh, founder of Astral and creator of uv, to dig into all of it.

Episode sponsors

Sentry Error Monitoring, Code talkpython26
Temporal
Talk Python Courses

Links from the show

Guests
Charlie Marsh: github.com
Ralf Gommers: github.com
Jonathan Dekhtiar: github.com

CPU dispatcher: numpy.org
build options: numpy.org
Red Hat RHEL: www.redhat.com
Red Hat RHEL AI: www.redhat.com
RedHats presentation: wheelnext.dev
CUDA release: developer.nvidia.com
requires a PEP: discuss.python.org
WheelNext: wheelnext.dev
Github repo: github.com
PEP 817: peps.python.org
PEP 825: discuss.python.org
uv: docs.astral.sh
A variant-enabled build of uv: astral.sh
pyx: astral.sh
pypackaging-native: pypackaging-native.github.io
PEP 784: peps.python.org

Watch this episode on YouTube: youtube.com
Episode #544 deep-dive: talkpython.fm/544
Episode transcripts: talkpython.fm

Theme Song: Developer Rap
🥁 Served in a Flask 🎸: talkpython.fm/flasksong

---== Don't be a stranger ==---
YouTube: youtube.com/@talkpython

Bluesky: @talkpython.fm
Mastodon: @talkpython@fosstodon.org
X.com: @talkpython

Michael on Bluesky: @mkennedy.codes
Michael on Mastodon: @mkennedy@fosstodon.org
Michael on X.com: @mkennedy