---
id: jax-cuda12-pjrt
version: "0.11.0"
license: Apache-2.0
license_treatment: permissive
maintenance: active
---

# jax-cuda12-pjrt — JAX XLA PJRT Plugin for NVIDIA GPUs

License: permissive · Maintenance: active · Downloads: 774.4K/mo

## What it is and what it does

jax-cuda12-pjrt is a plugin that connects JAX's XLA compiler to NVIDIA GPUs via the PJRT (Portable Runtime) interface, enabling GPU-accelerated execution of JAX programs on CUDA 12 hardware. It allows you to write numerical code using JAX's NumPy-like API and automatic differentiation, then compile and run it efficiently on NVIDIA accelerators.

The package is a specialized distribution component of JAX itself—it bundles the PJRT runtime plugin for CUDA 12 in a platform-specific wheel. Installation brings GPU support to JAX's core transformations (grad, jit, vmap) without requiring manual compilation of the plugin. It is actively maintained and carries no known security vulnerabilities.

Use it for:

- Training deep learning models on NVIDIA GPUs using JAX's automatic differentiation and JIT compilation.
- Running large-scale numerical simulations and scientific computing workloads with GPU acceleration.
- Developing machine learning research code that needs efficient gradient computation across GPU clusters.
- Accelerating array operations and linear algebra computations on NVIDIA hardware without writing CUDA code.
- Deploying JAX-based inference pipelines that require GPU performance for production inference.

## Worth the install?

AI-flagged interpretation of the facts above — verify before relying.

Provides NVIDIA GPU acceleration for JAX numerical computing via the PJRT plugin, enabling compiled array operations and automatic differentiation on CUDA 12 hardware.

Yes, if you are using JAX on an NVIDIA GPU with CUDA 12 and need GPU acceleration. The plugin is actively maintained, carries no known vulnerabilities, and is essential for GPU-backed JAX workloads. Install friction is moderate due to platform specificity and CUDA 12 runtime dependency, but this is expected for GPU-accelerated packages. Not needed if you are using CPU-only JAX or a different GPU backend.

## Install

pip install jax-cuda12-pjrt
uv add jax-cuda12-pjrt
poetry add jax-cuda12-pjrt

## Installing jax-cuda12-pjrt

Before you install: Medium install friction due to platform-specific wheels (manylinux_2_27 x86_64 and aarch64 only) and CUDA 12 runtime dependency. Active maintenance with recent releases; no known vulnerabilities.

License in practice: Apache-2.0 permissive license allows commercial and private use with minimal restrictions; suitable for most production and research deployments.

Quickstart:

pip install jax-cuda12-pjrt

import jax
import jax.numpy as jnp

# JAX operations now run on NVIDIA GPU
result = jax.jit(lambda x: jnp.dot(x, x))(jnp.ones((100, 100)))

Requires NVIDIA GPU with CUDA 12 support and CUDA 12 runtime libraries installed on the system; only available for Linux x86_64 and aarch64 platforms.

Verify before relying:

- Whether this package is a standalone plugin or requires a separate JAX installation as a prerequisite.
- Specific NVIDIA GPU compute capability requirements beyond CUDA 12 support.
- Performance characteristics compared to other JAX GPU backends or configurations.

## Package facts

- License: Apache-2.0 (permissive)
- Python support: unspecified
- Install friction: medium
- Maintenance: active
- Downloads: 774.4K/month (top 15,000 on PyPI)
- Known vulnerabilities: none known

## Tags

jax nvidia gpu acceleration, pjrt cuda 12 plugin, jax gpu backend, cuda accelerated numerical computing, jax xla gpu compilation, nvidia pjrt runtime, jax cuda support, gpu-acceleration, jax-plugin, cuda

[View on SkillFed](https://skillfed.io/packages/jax-cuda12-pjrt) · [View on PyPI](https://pypi.org/project/jax-cuda12-pjrt/)