The New Python 3.13 Is FINALLY Here!

Python 3.13 is here with exciting updates! In this video, I’ll dive into the key new features, including the game-changing option to disable the Global Interpreter Lock (GIL).

🔖 Chapters:
0:00 Intro
0:26 Improved Interactive Interpreter
3:33 Packages and modules
8:02 Deprecated decorator
8:46 _static_attributes_
10:34 Incremental Garbage Collection
14:13 Experimental JIT compiler
16:34 Support For iOS
16:56 Free-threaded Cpython
20:02 Final thoughts

#arjancodes #softwaredesign #python

[Gumitto]

20 minutes of watching your video saved me 5 minutes looking at python's release notes. It was worth it, thanks Arjan!

[rohitkochikkatfrancis]

You're the only reason my python is top quality thank you ....please don't stop.
Your channel is a gem❤❤

[nickhodgskin]

Yes, it's true that if you want to write performant code then using C or Rust is better. However just because you're writing Python doesn't mean performance isn't important at all. Python is such a widely used language, and is run at scale. Any performance improvements to CPython will have knock on effects to all Python code and reduce energy consumption and wasted computation on a global scale

[vaibhavarora]

As a python dev from almost a decade, i really enjoy your videos man . Keep em coming !!

[BrianStDenis-pjtq]

For us, its true we don't need multi-threaded Python very often. However, we just completed a project where we used an asynchronous design which had to contain several worker type functions that each had to be started as a task (to work with our async core). This works because the project does mostly I/O, even though one task does compress data. If we need more throughput, we'll get blocked by compressing and need more CPU power and the free thread improvement will then be beneficial. There has been and still are good reasons to use async, threaded and multiprocess application designs.

[MMYSLF]

Love your videos! Been a subscriber for a couple years. Appreciate all that you do to keep pythonistas informed!

If it's not too much trouble, cropping your videos to keep shaking imacs and macbooks out of the frame would be less distracting. I do enjoy seeing your nice office space though!

[ahmed-samer]

- great video, thank youu
- the garbage collection part is insightful but I think we need more details (could be in a separate video)
- the JIT section needed at least one working example using the `--enable-experimental-jit` flag
great effort as usual, keep it up

[wizardfix]

Good rundown, thanks. Just FYI, you never needed `quit()`, from the Python shell, as Ctrl+D quits instantly if the current line is empty.

[fytubevw]

Big hand for the team as well as this excellent intro!!

[kenneth.topp.]

the incremental GC was removed in 3.13.0rc3 (check the 3.13 changelog for details) due to a performance regression the sphinx team found.

[tjeerdhes]

Hey, I saw this guy yesterday as a speaker at the dutch pycon!

[JorgeEscobarMX]

I only read the free-threaded mode (PEP 703), and a Just-In-Time compiler (PEP 744) in depth. IS great to see they try to fix float point arithmetic with fma()

[taufiqurrahmansagafkelrey]

liked & shared!. you and your team ate!, no crumbs left, sir!!!! Thankyou for all of your videos.

[pablogonzalez]

It woud be great a video about running C/C++ code in python

[prashanthb]

I am waiting for JIT since ages.
I am happy Python has taken a good direction.

[borjasainz]

Oh Arjan! Today I will be in Utrecht 💫

[artistpw]

Thanks so much for this video. The new and updated features are great. But, truly can't wait for pi-thon.

[PanduPoluan]

I'm much more interested in the GC and JIT aspects of Python 3.13. GIL removal is not really important for me. Based on some benchmarks already floating around it seems that these 2 improvements indeed provide some observable performance over 3.12.

[wizardfix]

​ @ArjanCodes I think you're doing great work @ArjanCodes. You always have something interesting to tell us and you do it in a friendly and accessible way, informed by your experience. I (and, I expect, many others) would love it if you would make a video about how you produce these videos. The lowdown on recording hardware, editing, lighting, audio and video formats, software used, propitiating the Youtube algorithm etc., the step-by-step workflow; a metavideo if you like... I'd say many of your subscribers would be fascinated to learn about your process in this respect.

[vladimirtchuiev]

I would imagine free threaded could have performance increases in machine learning, maybe when moving data to and from the GPU, i would guess you would still have to test whether it would be better just to set a high number of workers in Pytorch...