A language and compiler for writing highly efficient custom deep-learning GPU kernels in Python-like code, with higher productivity than CUDA.
20.5K+35 this week3.4KMLIRMITWell maintainedUpdated 7 hours ago
2 alternatives
2 best open-source alternatives to CUDA (2026)
Looking for a free, open-source alternative to CUDA? These 2 projects can replace it. Each one passed our quality screening and is explained in plain words, so you can compare stars, license and maintenance at a glance.
Quick comparison
| Project | Stars | License | Last commit |
|---|---|---|---|
| triton-lang/triton | ★ 20.5K | MIT | 7 hours ago |
| KomputeProject/kompute | ★ 2.6K | Apache-2.0 | 3 months ago |
A C++ and Python GPU compute framework built on Vulkan that runs general-purpose workloads on AMD, Qualcomm, NVIDIA and other vendors' GPUs.
2.6K+2 this week202C++Apache-2.0Updated 3 months ago
FAQ
- Is there a free, open-source alternative to CUDA?
- Yes. The most popular open-source alternatives to CUDA on Repolode are triton, kompute.
- Which CUDA alternative should I choose?
- triton has the most GitHub stars and is released under MIT. Compare the license and last commit of each option in the table above before choosing.
How this list is made: AI reads each project's documentation to match it with products it can replace, every project passes our quality screening, and the list is ranked by GitHub stars. Stars updated Nov 7, 2026. How Repolode works