array-api-extra

What it is and what it does

array-api-extra extends the array API standard with additional utility functions, allowing developers to write array code that works across multiple backends (NumPy, CuPy, JAX, etc.) without lock-in to a single library. It sits on top of array-api-compat and is designed for libraries that need portable array operations.

The package is actively maintained and used by major scientific Python projects including SciPy, scikit-learn, and others. It requires Python 3.11 or later and carries an MIT license, making it safe to depend on in both open and closed projects.

Use it for:

• Build a library that works with multiple array backends without importing NumPy directly.
• Migrate numerical code from NumPy to a backend-agnostic implementation.
• Extend array-api-compat with domain-specific array operations for scientific computing.
• Develop GPU-accelerated code that can fall back to CPU arrays transparently.
• Create array utilities for machine learning frameworks that support multiple backends.

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

Provides extra array manipulation functions that work across any array library implementing the array API standard, enabling code that is not tied to NumPy or any single array backend.

Yes. The package is actively maintained, has low install friction, carries a permissive MIT license, and is already trusted by major projects like SciPy and scikit-learn. Install it when you need array operations that work across multiple backends or are building a library that should not depend on NumPy directly.

Install

array-api-extra on PyPI

pip install array-api-extra

uv add array-api-extra

poetry add array-api-extra

Installing array-api-extra

Before you install

Low friction install with a single runtime dependency (array-api-compat). Active maintenance with a release 2 days old as of the fact sheet date, and support for current Python versions (3.11 through 3.14).

License in practice

MIT license permits commercial and private use with minimal restrictions, making it suitable for integration into proprietary or open-source projects without licensing friction.

Quickstart

pip install array-api-extra

import array_api_extra as aae
# Use aae functions with any array-api-compliant array library

Requires Python 3.11 or later; works only with array libraries that implement the array API standard.

Verify before relying

• Which specific array functions are provided beyond what array-api-compat offers.
• Performance characteristics or overhead compared to direct NumPy calls.
• Scope of array API standard coverage (full or partial).

Package facts

Evidence: array_api_extra-0.11.1-py3-none-any.whl

Tags

More Scientific/Engineering packages