nvidia-nccl-cu12

NVIDIA Collective Communication Library (NCCL) Runtime

nvidia-nccl-cu12 License unclear Active v2.31.2 released 2026-08-11

Install

nvidia-nccl-cu12 on PyPI

pip

pip install nvidia-nccl-cu12

uv

uv add nvidia-nccl-cu12

poetry

poetry add nvidia-nccl-cu12

Package facts

License	not declared (unclear)
Python support	supports the current Python release (>=3)
Install friction	medium — platform-specific wheel
Runtime dependencies	none
Maintenance	actively maintained — 2 days since the last release
First released	2022-11-16
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: nvidia_nccl_cu12-2.31.2-py3-none-manylinux_2_18_aarch64.whl; nvidia_nccl_cu12-2.31.2-py3-none-manylinux_2_18_x86_64.whl

Keywords: cuda, nvidia, runtime, machine learning, deep learning

Development Status :: 4 - BetaIntended Audience :: DevelopersIntended Audience :: EducationIntended Audience :: Science/ResearchNatural Language :: EnglishOperating System :: Microsoft :: WindowsOperating System :: POSIX :: LinuxProgramming Language :: Python :: 3Programming Language :: Python :: 3 :: OnlyProgramming Language :: Python :: 3.10Programming Language :: Python :: 3.11Programming Language :: Python :: 3.5Programming Language :: Python :: 3.6Programming Language :: Python :: 3.7Programming Language :: Python :: 3.8Programming Language :: Python :: 3.9Topic :: Scientific/EngineeringTopic :: Scientific/Engineering :: Artificial IntelligenceTopic :: Scientific/Engineering :: MathematicsTopic :: Software DevelopmentTopic :: Software Development :: Libraries

About nvidia-nccl-cu12

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

NCCL (pronounced "Nickel") is a stand-alone library of standard collective communication routines for GPUs, implementing all-reduce, all-gather, reduce, broadcast, and reduce-scatter. It has been optimized to achieve high bandwidth on any platform using PCIe, NVLink, NVswitch, as well as networking using InfiniBand Verbs or TCP/IP sockets.

Read as markdown · JSON record · Homepage

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

NCCL provides GPU collective communication primitives (all-reduce, all-gather, reduce, broadcast, reduce-scatter) optimized for high-bandwidth transfers across PCIe, NVLink, NVswitch, InfiniBand, and TCP/IP.

Medium install friction due to platform-specific wheels (aarch64 and x86_64 manylinux only); actively maintained with a release 2 days old, indicating current support.

License status is unclear—no SPDX identifier or raw license text is available in the package metadata, so users should verify licensing terms directly with NVIDIA before adoption.

Usage

pip install nvidia-nccl-cu12==2.31.2

Requires CUDA 12.x runtime and GPU hardware; wheels are only available for Linux aarch64 and x86_64 architectures.

Verdict: Active NVIDIA library for GPU collective communication, recently updated and widely used (top 1000 PyPI). Medium install friction due to platform specificity; no known vulnerabilities. Primary concern is unclear licensing—verify terms before production use.

Needs verification

Whether the package includes or requires separate NVIDIA NCCL C library installation beyond the wheel
Exact license terms and any restrictions on commercial or research use
How collective communication primitives are exposed to Python callers

gpu collective communicationnccl nvidiadistributed gpu trainingall-reduce all-gather gpumulti-gpu synchronizationcuda collective operationsgpu communication library