Officially:
* the Numworks calculator uses MicroPython, the first official implementation of Python on a graphing calculator, made in several weeks time, and certainly not full-time;
* on the Graph 90+E / fx-CG50, Casio's using MicroPython as well;
* the HP Prime uses modified snapshots of giac, in which Bernard Parisse recently implemented quite good compatibility with the Python syntax. That's not real Python, but it doesn't make
too much of a difference for the intended use case: simple algorithms for school usage.
Unofficially:
* on the Nspire, in 2014, Vogtinator used the lightweight MicroPython implementation. He didn't spend weeks on it either;
* Bernard Parisse also ported giac to the Nspire series years ago (KhiCAS), and recently to Graph 90+E / fx-CG50, so new builds of giac for these platforms provide a Python-like language;
* "zardam" ported giac to a modified NumWorks calculator featuring an extra Flash memory chip adding 8 or 16 MB of storage, so in a sense, there can be two different ways to access a Python(-like) language on the NumWorks platform.
Down the road, the first NumWorks calculator model will suffer from its limited hardware (if kept in its pristine state, that is). They bought their large batch of SoCs slightly before ST introduced new models with more RAM and most of all, more Flash memory, so out of the box, the NumWorks calculators are stuck on 256 KB of RAM and 1 MB of Flash...
On the flip side, the PCB of all calculators of the first model has the pads and wires for adding an external NOR Flash chip of 8 MB for less than 1 USD / EUR, or 16 MB for less than 3 USD / EUR; the pitch is 1.27mm (0.05"), so the chip can be soldered without high-end equipment or skills. I've never soldered pitches below 2 mm myself, but I know from other people that 1.27mm is tolerable with a non-damaged soldering iron tip.
I don't know whether, and when, NumWorks plans on releasing another model containing another STM32F4 chip featuring more RAM and Flash memory