Overview

NumPy 1.25, released on June 18, 2023, improves Array API standard compatibility and continues type modernization.

Main Features

Improved Array API

NumPy 1.25 progresses toward Array API standard compliance, facilitating interoperability with CuPy, JAX, and other numerical computing libraries.

python 
import numpy as np

# Array API namespace
xp = np  # Array API compatible

# Portable code across NumPy, CuPy, JAX
def normalize(x):
    xp = x.__array_namespace__()
    return (x - xp.mean(x)) / xp.std(x)

data = np.array([1.0, 2.0, 3.0, 4.0, 5.0])
print(normalize(data))

Type modernization

Type promotion handling has been improved to be more predictable, and new deprecation warnings prepare for version 2.0 changes.

python 
import numpy as np

# More predictable type promotion
a = np.array([1, 2, 3], dtype=np.int32)
b = np.array([1.0, 2.0, 3.0], dtype=np.float32)

# Result is float64 (clear promotion)
c = a + b
print(c.dtype)  # float64

# New: explicit control
d = np.add(a, b, dtype=np.float32)
print(d.dtype)  # float32

Sources