ml-dtypes

ml_dtypes is a stand-alone implementation of several NumPy dtype extensions used in machine learning.

ml-dtypes Permissive license Apache-2.0 Active ★ 357 v0.6.0 released 2026-08-13

Install

ml-dtypes on PyPI

pip

pip install ml-dtypes

uv

uv add ml-dtypes

poetry

poetry add ml-dtypes

Package facts

License	Apache-2.0 (permissive)
Python support	supports the current Python release (>=3.10)
Install friction	medium — platform-specific wheel
Runtime dependencies	1 — numpy
Maintenance	actively maintained — 0 days since the last release
Last repo commit	2026-08-13
First released	2023-03-10
Popularity	one of the top 1,000 most-downloaded packages on PyPI (30-day window, as of 2026-08-13)
Known vulnerabilities	none known (OSV.dev, checked 2026-08-13)

Evidence: ml_dtypes-0.6.0-cp310-cp310-macosx_10_9_universal2.whl; ml_dtypes-0.6.0-cp310-cp310-manylinux_2_27_aarch64.manylinux_2_28_aarch64.whl; ml_dtypes-0.6.0-cp310-cp310-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl; ml_dtypes-0.6.0-cp310-cp310-win_amd64.whl; ml_dtypes-0.6.0-cp311-cp311-macosx_10_9_universal2.whl; ml_dtypes-0.6.0-cp311-cp311-manylinux_2_27_aarch64.manylinux_2_28_aarch64.whl; ml_dtypes-0.6.0-cp311-cp311-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl; ml_dtypes-0.6.0-cp311-cp311-win_amd64.whl; ml_dtypes-0.6.0-cp311-cp311-win_arm64.whl; ml_dtypes-0.6.0-cp312-cp312-macosx_10_13_universal2.whl; ml_dtypes-0.6.0-cp312-cp312-manylinux_2_27_aarch64.manylinux_2_28_aarch64.whl; ml_dtypes-0.6.0-cp312-cp312-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl; ml_dtypes-0.6.0-cp312-cp312-win_amd64.whl; ml_dtypes-0.6.0-cp312-cp312-win_arm64.whl; ml_dtypes-0.6.0-cp313-cp313-macosx_10_13_universal2.whl; ml_dtypes-0.6.0-cp313-cp313-manylinux_2_27_aarch64.manylinux_2_28_aarch64.whl; ml_dtypes-0.6.0-cp313-cp313-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl; ml_dtypes-0.6.0-cp313-cp313-win_amd64.whl; ml_dtypes-0.6.0-cp313-cp313-win_arm64.whl; ml_dtypes-0.6.0-cp314-cp314-macosx_10_15_universal2.whl

Intended Audience :: Science/ResearchProgramming Language :: Python :: 3Programming Language :: Python :: 3 :: Only

About ml-dtypes

from the package's own PyPI description — quoted content, verbatim

ml_dtypes

Unittests (image) Wheel Build (image) PyPI version (image)

ml_dtypes is a stand-alone implementation of several NumPy dtype extensions used in machine learning libraries, including:

bfloat16: an alternative to the standard float16 format
8-bit floating point representations, parameterized by number of exponent and mantissa bits, as well as the bias (if any) and representability of infinity, NaN, and signed zero.
float8_e3m4
float8_e4m3
float8_e4m3b11fnuz
float8_e4m3fn
float8_e4m3fnuz
float8_e5m2
float8_e5m2fnuz
float8_e8m0fnu
Microscaling (MX) sub-byte floating point representations:
float4_e2m1fn
float6_e2m3fn *...

Read as markdown · JSON record · Source repository

AI interpretation — verify before relying

AI-generated interpretation of the package facts above; every digit, version, license, or vulnerability id it cites is grounded in the facts already shown on this page

ml_dtypes provides NumPy-compatible data types for machine learning, including bfloat16, 8-bit floats (float8_e5m2, float8_e4m3, etc.), 4–6-bit microscaling formats, and narrow integers (int1–int4, uint1–uint4).

Medium install friction due to platform-specific wheels (cp310–cp314 across macOS, Linux, Windows, and ARM64), but pre-built binaries are available for all major platforms. Active maintenance with a recent release.

Apache-2.0 permissive license; pre-compiled wheels include EIGEN (MPL 2.0). Both are compatible with commercial use, though MPL 2.0 requires source disclosure for modifications to EIGEN itself.

Usage

pip install ml_dtypes
from ml_dtypes import bfloat16
import numpy as np
arr = np.zeros(4, dtype=bfloat16)

Requires Python >=3.10; platform-specific wheel selection is automatic but may require a recent pip/setuptools.

Verdict: A well-maintained, actively developed library for low-precision ML dtypes with broad platform support and no known vulnerabilities. Apache-2.0 licensing is permissive. Medium install friction is typical for compiled extensions; suitable for ML workflows that need bfloat16 or float8 support in NumPy.

Needs verification

Whether numpy's aggregation behavior with these dtypes (e.g., precision loss in sum) is documented or mitigated in recent versions.
Performance characteristics and overhead of unpacked sub-byte integer types relative to standard NumPy integers.

bfloat16 numpy dtypefloat8 machine learninglow precision data typesml_dtypes numpy extensions8-bit floating pointmicroscaling formatsnarrow integer types