nvidia-cuda-nvrtc-cu11

What it is and what it does

This package bundles NVIDIA's NVRTC (CUDA Runtime Compiler) native libraries as a Python wheel, allowing developers to compile CUDA kernels at runtime rather than ahead-of-time. It is a low-level distribution mechanism for CUDA 11 tooling, not a high-level API—it exists primarily to satisfy dependencies of higher-level frameworks like CuPy or other GPU acceleration libraries that need runtime kernel compilation capabilities.

The package provides no Python API of its own; it simply makes compiled NVIDIA libraries available to other packages. Installation requires matching your system architecture (x86_64 or aarch64 Linux, or x86_64 Windows) and Python version (3.5 through 3.11). Because the package is abandoned and has not been updated since October 2022, it is locked to CUDA 11.8.89 and will not receive updates for newer CUDA versions or security patches.

Use it for:

• Satisfy NVRTC dependencies for GPU-accelerated Python libraries like CuPy that perform runtime kernel compilation
• Enable just-in-time CUDA kernel compilation in scientific computing workflows without pre-compiling kernels
• Support legacy projects pinned to CUDA 11 that require runtime compilation capabilities
• Provide NVRTC libraries in containerized environments where CUDA toolkit installation is not feasible

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

Provides NVIDIA CUDA NVRTC (runtime compilation) native libraries for Python, enabling just-in-time compilation of CUDA kernels on x86_64 and aarch64 Linux and Windows systems.

Yes, but only if you need CUDA 11.8.89 NVRTC support and have no alternative. The package is abandoned and will not receive updates; if you are starting a new project, prefer a maintained CUDA distribution or newer CUDA version. Install only as a transitive dependency of a package you already require, or if you are locked to CUDA 11 by other constraints. Verify NVIDIA's proprietary license permits your use case before deployment.

Install

nvidia-cuda-nvrtc-cu11 on PyPI

pip install nvidia-cuda-nvrtc-cu11

uv add nvidia-cuda-nvrtc-cu11

poetry add nvidia-cuda-nvrtc-cu11

Installing nvidia-cuda-nvrtc-cu11

Before you install

Medium install friction due to platform-specific wheel distributions (manylinux1, manylinux2014, Windows). Package is abandoned as of 1411 days since last release, with no active maintenance or repository commits tracked.

License in practice

Licensed under NVIDIA Proprietary Software with unclear license treatment. Users should review NVIDIA's licensing terms before deployment, particularly for commercial or redistributed applications.

Quickstart

pip install nvidia-cuda-nvrtc-cu11==11.8.89

import nvidia.cuda.nvrtc_cu11
# Access NVRTC runtime libraries for kernel compilation

Requires CUDA 11 compatible system; only available for Python 3.5+ on Linux (x86_64/aarch64) or Windows (x86_64).

Verify before relying

• Whether abandoned status affects compatibility with current CUDA toolchain versions beyond 11.8.89
• Whether proprietary license permits use in open-source or commercial projects without explicit NVIDIA agreement
• Current stability and security posture given 1411 days without maintenance

Package facts

Evidence: nvidia_cuda_nvrtc_cu11-11.8.89-py3-none-manylinux1_x86_64.whl; nvidia_cuda_nvrtc_cu11-11.8.89-py3-none-manylinux2014_aarch64.whl; nvidia_cuda_nvrtc_cu11-11.8.89-py3-none-manylinux2014_x86_64.whl; nvidia_cuda_nvrtc_cu11-11.8.89-py3-none-win_amd64.whl

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

Tags

More Software Development packages

Further reading