Each instance normally keeps its attributes in a dict, obj.__dict__. That is flexible, and it costs memory for every object. Listing the attribute names in __slots__ gives instances fixed storage instead.

class Plain:
    def __init__(self, x):
        self.x = x

class Slim:
    __slots__ = ("x",)

    def __init__(self, x):
        self.x = x

p, s = Plain(1), Slim(1)
p.typo = 2  # silently accepted
print(hasattr(s, "__dict__"))
try:
    s.typo = 2
except AttributeError as e:
    print(e)

• Assigning a name that is not in __slots__ raises AttributeError, which catches misspellings such as self.raduis.
• Instances are smaller and attribute access is a little faster. It matters when you create many thousands of objects.
• A subclass without its own __slots__ gets a __dict__ back. Give each subclass __slots__ listing only its new names.
• A slotted name cannot also have a class-level default. Set defaults in __init__.

@dataclass(slots=True) writes __slots__ for you.

Give Vector(x, y) __slots__ so instances have no __dict__ and reject unknown attributes. Make Vector3(Vector), which adds z defaulting to 0, keep the same restriction by listing only its new name in its own __slots__.

The explanation opens here once every check passes.