nixl
NIXL Python API meta package for CUDA variants
Decision gist · record as of 2026-08-14
Yes, if you use PyTorch with CUDA and want to avoid manual backend selection. The package is actively maintained, has no known vulnerabilities, uses permissive licensing, and installs with low friction. Install it when you need transparent CUDA version detection; skip it if you manually manage CUDA backends or use CPU-only PyTorch.AI-flagged interpretation of the facts on this page — verify before relying
Before you install
- Requires Python 3.10 or later and a PyTorch installation with CUDA support already configured on your system.
- Active maintenance with a recent release.
- Low install friction as a pure Python wheel.
License · maintenance · safety
MIT AND Apache-2.0 (permissive) — Licensed under MIT AND Apache-2.0 (permissive dual licensing), allowing use in most commercial and open-source projects without restriction.
last release 2026-08-14 (0 days)
0 known vulnerabilities (OSV.dev, 2026-08-14) · 500,426 downloads/mo, #6,320 on PyPI
Alternatives
Verify before relying
pip install nixl
import nixl
# Backend selection happens automatically at import time based on system CUDA version- Whether the automatic CUDA backend selection works correctly across all PyTorch versions and CUDA driver configurations
- Performance characteristics or overhead of the runtime backend selection mechanism
- Compatibility with non-PyTorch CUDA applications or mixed CUDA environments
What it is and what it does
nixl is a meta package that bundles both CUDA 12 and CUDA 13 backends for PyTorch. When you install it via pip, you get both backends included, and the package automatically detects your system's CUDA version at runtime to select the correct one. This eliminates the need to manually choose and install the right CUDA variant yourself. The package is actively maintained and released recently, with low installation friction since it distributes as a pure Python wheel.
The design is particularly useful for environments where CUDA versions might vary or where you want to distribute a single package that works across different CUDA setups. The `nixl[cu12]` and `nixl[cu13]` extras are supported for backward compatibility but don't change the behavior—the automatic selection still occurs at runtime.
Use it for
- Deploying PyTorch applications across systems with different CUDA versions without rebuilding or maintaining separate packages
- Simplifying dependency management in containerized environments where CUDA version detection is needed at runtime
- Building Python packages that depend on nixl and need to work with multiple CUDA backends transparently
- Testing PyTorch code against both CUDA 12 and CUDA 13 without manual backend switching
Worth the install?
AI-flagged interpretation of the facts on this page. Verify before relying on it.
Yes, if you use PyTorch with CUDA and want to avoid manual backend selection.
The package is actively maintained, has no known vulnerabilities, uses permissive licensing, and installs with low friction. Install it when you need transparent CUDA version detection; skip it if you manually manage CUDA backends or use CPU-only PyTorch.
Install
nixl on PyPI
Before you install
Active maintenance with a recent release. Low install friction as a pure Python wheel. Requires Python 3.10 or later.
Requires Python 3.10 or later and a PyTorch installation with CUDA support already configured on your system.
License in practice
Licensed under MIT AND Apache-2.0 (permissive dual licensing), allowing use in most commercial and open-source projects without restriction.
Quickstart
pip install nixl
import nixl
# Backend selection happens automatically at import time based on system CUDA versionVerify before relying
- Whether the automatic CUDA backend selection works correctly across all PyTorch versions and CUDA driver configurations
- Performance characteristics or overhead of the runtime backend selection mechanism
- Compatibility with non-PyTorch CUDA applications or mixed CUDA environments
Package facts
| License | MIT AND Apache-2.0 permissive |
| Python support | Supports the current Python release >=3.10 |
| Install friction | Low. Pure-Python wheel |
| Runtime dependencies | 2 packagesnixl-cu12nixl-cu13 |
| Maintenance | Actively maintained 0 days since the last release |
| First released | |
| Downloads | 500,426 / month, #6,320 on PyPI 30-day window, as of 2026-08-14 |
| Known vulnerabilities | None known OSV.dev, checked 2026-08-14 |
Evidence: nixl-1.4.0-py3-none-any.whl
Tags
Let your AI agent find packages like this
Example. Real query, live index.
You found this page by searching. An agent finds it by wishing: SkillFed indexes 14,416 PyPI packages by what they can do, searchable in plain language.
wish › “cuda backend selector”
- nixlnixl is a meta package that automatically selects and installs the…
- nvidia-nvvmProvides the NVVM compiler IR library for building and optimizing…
- jax-cuda12-pluginEnables JAX to run numerical computations and machine learning…
Give your agent the search over MCP, or paste the wish link into any chat.
More Artificial Intelligence packages
LiteLLM provides a unified Python interface to call 100+ LLM providers (OpenAI, Anthropic, Gemini, Bedrock, Azure, and others) using OpenAI-compatible API format, available as both a Python SDK and a self-hosted AI Gateway proxy server.
Install it if you need to work with multiple LLM providers or want to centralize LLM routing in your organization.
Client library and CLI tool for downloading, uploading, and managing models, datasets, and repositories on the Hugging Face Hub platform.
Install it if you work with Hugging Face Hub models or datasets.
LangChain provides a framework for building agents and LLM-powered applications by composing language models, tools, and memory through a unified API that abstracts over multiple model providers.
hf-xet provides chunk-based deduplication and efficient file transfer for the Hugging Face Hub, enabling faster uploads and downloads of large files with local disk caching.
Tokenizers converts raw text into token sequences for NLP models, with support for training custom vocabularies and using pre-built tokenizers (BPE, WordPiece) optimized for speed via Rust.
Transformers provides a unified framework for loading, fine-tuning, and running state-of-the-art pretrained models across text, vision, audio, video, and multimodal tasks using PyTorch, JAX, or TensorFlow.
Install it if you need to run or train any transformer-based model for NLP, vision, audio, or multimodal tasks.
See also nixl-cu13 · nixl-cu12 · cuda-toolkit · nvidia-cusparse-cu12 · nvidia-cudnn-cu13 · nccl4py · slangtorch · cuequivariance-torch · cuda-core · flash-attn-4