You can use `uv python install 3.13 --reinstall` and we'll replace the existing one. Virtual environments just contain a reference to the distribution.

uv makes it easy to benchmark uv

These numbers are for CPython 3.13.2.

The benchmark is a simple one — pulled from a report that the conda-forge distribution was more performant than ours. (Now, we're looking ~20% faster than conda-forge on macOS.)

The benchmark is available at gist.github.com/zanieb/551ed...

And on x86-64 Linux

For more on the new uv release, see github.com/astral-sh/uv...

For more on the tail calling interpreter, see docs.python.org/3.14/whatsne... and github.com/python/cpyth...

1.23x just is the geometric mean on the pyperformance benchmark suite. There are more significant speedups for some workloads!

Here's a full list, gist.github.com/zanieb/84748...

In the latest version, `uvx --with pyyaml python` is supported (which reads a little more clearly and can be repeated to include more dependencies)

Thanks for sharing it! We look at those every once and a while. I wish we could break it down into monthly amounts like the PyPI stats :)

I see that now 🤦‍♀️ haha

There was some recent discussion about this in github.com/python/cpyth...

I imagine there are other people interested in helping to upstream the changes. Maybe we can find a space to discuss collaboration on Discord?

I'm very interested in this! My next step was to find CPython developers willing to mentor the work — it's great to hear that you're willing to review the changes.

It's a bit of a struggle to find time, but we're investing more resources into this project and upstream support is motivating :)