---
id: ml-dtypes
version: "0.6.0"
license: Apache-2.0
license_treatment: permissive
maintenance: active
---

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

License: permissive · Maintenance: active · Popularity: top 1,000 on PyPI

## Install

pip install ml-dtypes
uv add ml-dtypes
poetry add ml-dtypes

## Description

# ml_dtypes

[![Unittests](https://github.com/jax-ml/ml_dtypes/actions/workflows/test.yml/badge.svg)](https://github.com/jax-ml/ml_dtypes/actions/workflows/test.yml)
[![Wheel Build](https://github.com/jax-ml/ml_dtypes/actions/workflows/wheels.yml/badge.svg)](https://github.com/jax-ml/ml_dtypes/actions/workflows/wheels.yml)
[![PyPI version](https://badge.fury.io/py/ml_dtypes.svg)](https://badge.fury.io/py/ml_dtypes)

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

- [`bfloat16`](https://en.wikipedia.org/wiki/Bfloat16_floating-point_format):
  an alternative to the standard [`float16`](https://en.wikipedia.org/wiki/Half-precision_floating-point_format) 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`
  *...

## AI interpretation — verify before relying

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).

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.

[View on SkillFed](https://skillfed.io/packages/ml-dtypes) · [View on PyPI](https://pypi.org/project/ml-dtypes/)