I can reproduce this on the main branch, but only with mypy:

import numpy as np
import numpy.typing as npt

a: npt.NDArray[np.float64]
reveal_type(iter(a))

Mypy 1.14.1:

Revealed type is "typing.Iterator[numpy.float64]"

Pyright 1.1.392:

Type of "iter(a)" is "Iterator[Any]"

This is caused by a bug in mypy (probably python/mypy#14070).

To see why I think mypy is behaving incorrectly, you could take a look at the definition for numpy.ndarray.__iter__:

The only overload that could return a Iterator[float64], is the second one. This overload requires self to have a shape-type of tuple[int], i.e. it must be 1-d. But the shape-type of a is unspecified, i.e. tuple[int, ...]. But tuple[int, ...] cannot be assigned to tuple[int].
So the only valid overload here, is the last one, which returns Iterator[Any], and is therefore what Pyright infers the return type as.

I don't see any reason why mypy would behave this way,.. So I also don't know how we could work around this. It's probably a good idea to raise this with mypy &mdhash; they have been very helpful recently in fixing numpy-related issues (see python/mypy#18343).

Thanks for taking the time to look into this and suggesting a possible cause of the problem. I've raised the issue with mypy: python/mypy#18566. I'll go ahead and close this issue and continue monitoring the situation on the mypy side.