# NVIDIA/cuda-python

CUDA Python: Performance meets Productivity

Repository: https://github.com/NVIDIA/cuda-python
Canonical: https://ross.abutalabs.com/products/cuda-python
Homepage: https://nvidia.github.io/cuda-python/
Language: Cython
License: Apache-2.0
License Family: permissive
Last push: 2026-08-26T21:46:14+00:00

## Health v2 (maintenance only)
Score: 95/100 (v2, computed 2026-09-03T02:39:23.370411+00:00)
- activity 99, release rhythm 87, longevity 100
- inputs: {"age_days": 1892, "days_push": 7, "days_rel": 9, "gap_med": 6.0, "n_releases_24m": 69}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 3359, forks 328 (observed 2026-08-28T04:07:57.869174+00:00)

## What it is
NVIDIA's official Python package for accessing the CUDA platform, providing Pythonic APIs and low-level bindings to CUDA C APIs, parallel algorithms, JIT kernel compilation, and math libraries. It is an umbrella project comprising components like cuda.core, cuda.bindings, numba.cuda, cuda.tile, and nvmath-python.

## Use cases
- write CUDA kernels in Python
- call CUDA C APIs from Python
- run parallel sort and reduce on the GPU
- compile Python functions to GPU code with Numba
- access cuBLAS and other NVIDIA math libraries from Python
- profile CUDA Python applications
- locate installed CUDA libraries in a Python environment

## When to choose
- you need low-level control over CUDA from Python
- you want to write GPU kernels without leaving Python
- you need Python bindings for NVIDIA math or communication libraries
- you are building GPU-accelerated Python libraries or frameworks

## When to avoid
- you only need high-level deep learning frameworks like PyTorch or TensorFlow
- you have no NVIDIA GPU or CUDA-capable hardware
- you need a pure-CPU numerical computing solution

## Facets
- artifact type: library
- maturity: active
- function: gpu-computing, sdk, compiler, math, concurrency
- domain: gpu-computing, developer-tools, machine-learning, data-science, performance
- platform: python, windows
- tags: cuda, nvidia, python-bindings, gpu-kernels, numba, jit-compilation, parallel-computing, cython, linux, macos, gpu

## Member repositories
- NVIDIA/cuda-python (main) score 95

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:07:57.869174+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T18:40:52.862546+00:00, confidence not recorded.
  - readme: https://github.com/NVIDIA/cuda-python (fetched 2026-08-28T04:07:57.869174+00:00, sha 4276dc0afa6c)
  - homepage: https://nvidia.github.io/cuda-python/ (fetched 2026-08-29T09:33:51.848600+00:00, sha cfea376851ca)
  - registry_pypi: https://pypi.org/pypi/cuda-python/json (fetched 2026-08-29T09:33:51.857856+00:00, sha 6510ceaa7f36)
- Data as of 2026-08-30T08:39:29.467469+00:00.