NVIDIA Collective Communication Library (NCCL) Runtime
Project Links
Meta
Author: Nvidia CUDA Installer Team
Requires Python: >=3
Classifiers
Development Status
- 4 - Beta
Intended Audience
- Developers
- Education
- Science/Research
Natural Language
- English
Programming Language
- Python :: 3
- Python :: 3.5
- Python :: 3.6
- Python :: 3.7
- Python :: 3.8
- Python :: 3.9
- Python :: 3.10
- Python :: 3.11
- Python :: 3 :: Only
Topic
- Scientific/Engineering
- Scientific/Engineering :: Mathematics
- Scientific/Engineering :: Artificial Intelligence
- Software Development
- Software Development :: Libraries
Operating System
- Microsoft :: Windows
- POSIX :: Linux
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.
2.28.7
Oct 21, 2025
2.28.3
Sep 06, 2025
2.27.7
Aug 04, 2025
2.27.6
Jul 15, 2025
2.27.5
Jun 26, 2025
2.27.3
Jun 03, 2025
2.26.5
May 02, 2025
2.26.2.post1
Apr 08, 2025
2.26.2
Mar 13, 2025
2.25.1.post1
Apr 08, 2025
2.25.1
Jan 24, 2025
2.24.3
Jan 07, 2025
2.23.4
Sep 11, 2024
2.22.3
Jun 19, 2024
2.21.5
Apr 03, 2024
2.20.5
Mar 06, 2024
2.20.3
Apr 02, 2024
2.19.3
Oct 24, 2023
2.18.3
Jun 13, 2023
2.18.1
May 04, 2023
2.17.1
Mar 31, 2023
2.16.5
Feb 02, 2023
2.16.2
Dec 09, 2022
0.0.1.dev5
Nov 16, 2022
Wheel compatibility matrix
Files in release
No dependencies