Version 0.64.0 (18 February 2026)

This is a major Numba release. Numba now supports NumPy 2.4.

Please find a summary of all noteworthy items below.

Highlights

Support for NumPy 2.4

Numba now supports NumPy 2.4, with the following notable changes:

  • np.trapz has been removed in NumPy 2.4. Use np.trapezoid instead (available since NumPy 2.0).

  • np.in1d has been removed in NumPy 2.4. Use np.isin instead.

(PR-#10393)

NumPy Support

Add support for np.moveaxis

Support is added for NumPy moveaxis function.

(PR-#10366)

Bug Fixes

Fix scalar handling in np.all

Fixed scalar handling in the np.all function. Previously, this function would fail when called with scalar inputs. Now it properly handles both scalar and array inputs, converting scalars to boolean values as expected.

(PR-#10223)

Fix scalar handling in np.any

Fix scalar handling in np.any function. Previously, this function would fail when called with scalar inputs. Now it properly handles both scalar and array inputs, converting scalars to boolean values as expected.

(PR-#10224)

Fix np.asfortranarray and np.ascontiguousarray readonly handling

Fix readonly input producing readonly output even when a copy is made using np.asfortranarray or np.ascontiguousarray.

(PR-#10390)

Support integer type promotion in Python API

Fixes a segmentation fault on s390x (IBM Z) by ensuring integer types are properly promoted to full register width in the CPU and Python API lowering layers, as required by the SystemZ ABI.

(PR-#10396)

Fix precision loss in integer power operations on Python 3.11

Fixed precision loss in integer power operations (e.g., x ** 2) for large integer values on Python 3.11. Integer results now preserve full precision for values greater than 2^53.

(PR-#10398)

Pull-Requests:

Authors: