vosen

ZLUDA: Run CUDA Apps on Non-NVIDIA GPUs

ZLUDA is a Rust-based CUDA replacement intended to run unmodified CUDA applications on non-NVIDIA GPUs. It targets users who need to run CUDA software without NVIDIA hardware; check the documentation for setup and compatibility details.

Stars
15.2k
Forks
969
Stars, last 30 days
+113
Last commit
6 days ago
Contributors
17

Our take

Activeregular commits and releases

Health score

50/100

How it is scored
Activity
71
Community
43
Issues
46
Pull requests
41

+113 stars in the last 30 days

  • Releases more often than 91% of the projects we track

ZLUDA is actively maintained and releases frequently, but adoption carries compatibility uncertainty and contributor-concentration risk.

Good fit if

  • You need to try unmodified CUDA applications on non-NVIDIA GPUs and can verify your specific hardware and workload.
  • You are comfortable evaluating an actively developed Rust project with frequent preview releases.
  • Your team can account for the Apache-2.0 license and assess whether its terms suit your use.

Look elsewhere if

  • You need confirmed support for a particular GPU or operating system that the documentation does not establish.
  • Your deployment requires a stable release rather than preview releases.
  • You need a project whose maintenance is not concentrated: one contributor accounts for most contributions.
All health signals
Last commit2026-10-26 (6 days ago)
Commits, last 90 days16
Releases, last 12 months88 (latest v7-preview.19, 2026-10-26, pre-release)
Contributors17 (top contributor: 77% of commits)
Issues closed, last 90 days1
Pull requests merged, last 90 days16 (typically merged in 2 days)
Project age6 years

Checked on 2026-11-02 with the GitHub API.

Project history

Daily GitHub snapshots since Oct 7, 2026. History covers stars and forks; health signals are not tracked historically.

Selected snapshot

Stars

15,216

Forks

974

Change since Nov 17, 2026: +8 stars·+1 forks

Overview

ZLUDA is a Rust-based drop-in replacement for CUDA that aims to run unmodified CUDA applications on non-NVIDIA GPUs. It addresses the gap between CUDA software and users whose hardware is not from NVIDIA, with the project describing its goal as near-native performance.

The project is for people who need to run existing CUDA applications on supported non-NVIDIA hardware. The supplied overview does not name specific GPU models or operating systems, so consult the documentation before relying on compatibility.

Key Features

  • Acts as a drop-in CUDA replacement for non-NVIDIA GPUs.
  • Aims to run unmodified CUDA applications.
  • Targets near-native performance, as stated by the project.
  • Provides setup guidance through its quick-start documentation.
  • Offers community discussion through its Discord server.
  • Publishes project updates on its news page.

Use Cases

  • Developers testing whether an existing CUDA application can run on non-NVIDIA GPU hardware.
  • Teams evaluating alternatives to NVIDIA hardware for a CUDA-dependent workload.
  • Users exploring whether their current GPU can support software written for CUDA, after confirming compatibility in the documentation.

What you need

Detected in the repository

  • Rust (from Cargo.toml)
  • Automated checks on GitHub Actions

License in plain words

Apache-2.0permissive

  • Commercial use: yes
  • Modify and redistribute: yes
  • You must keep: the license, the NOTICE file and a note of your changes
  • Share your changes: no
  • Includes an explicit patent grant from the contributors:

A summary, not legal advice: the LICENSE file is what applies.

Getting Started

Start with the quick-start documentation for installation and compatibility instructions. The repository README links to the same guide.

Considerations

The project is active, with regular commits and releases, including frequent preview releases. However, most contributions come from one person, and few issues have been closed recently, so teams should assess their support needs and monitor project progress. The supplied requirements identify Rust, but do not specify GPU models, supported operating systems, or a test suite; check the documentation for hardware and application compatibility before adoption.

Found this useful?

Share it with someone who would like ZLUDA.