Decision Workspace
teeny-cuda vs burn-cuda vs rlx-gpu-kernels
Side-by-side comparison of Rust crates
50
teeny-cuda
experimentalv0.2.0
Teenygrad is a high-performance, memory-safe Rust ML training and inference library. It targets devices from microcontrollers to traditional GPUs with statically-typed kernels and full async support.
74
burn-cuda
growingv0.22.0
CUDA backend for the Burn framework
64
rlx-gpu-kernels
experimentalv0.2.17
Shared CUDA/HIP C++ kernel sources for RLX GPU backends (NVRTC / hipRTC).
Core Metrics
| teeny-cuda | burn-cuda | rlx-gpu-kernels | |
|---|---|---|---|
| Health Score | 50 | 74 | 64 |
| Total Downloads | 242 | 1.3M | 3.9K |
| 30d Downloads | 0 | 0 | 0 |
| Dependents | 13 | 368 | 30 |
| Releases | 5 | 28 | 16 |
| Last Updated | 55d ago | 3d ago | 6d ago |
| Age | 2m | 2y 1m | 4m |
Health Breakdown
teeny-cuda
Maintenance
11
Quality
14
Community
9
Popularity
3
Documentation
13
burn-cuda
Maintenance
20
Quality
17
Community
17
Popularity
7
Documentation
13
rlx-gpu-kernels
Maintenance
20
Quality
15
Community
10
Popularity
4
Documentation
15
Technical Details
| teeny-cuda | burn-cuda | rlx-gpu-kernels | |
|---|---|---|---|
| Version | 0.2.0 | 0.22.0 | 0.2.17 |
| Stable (≥1.0) | ✗ No | ✗ No | ✗ No |
| License | Apache-2.0 | MIT OR Apache-2.0 | MIT OR Apache-2.0 |
| Dependencies | 8 | 4 | 3 |
| Crate Size | 56KB | 37KB | 143KB |
| Features | 2 | 7 | 3 |
| Yanked % | 0.0% | 0.0% | 0.0% |
| Edition | 2024 | 2024 | 2024 |
| MSRV | — | — | 1.89 |
| Owners | 1 | 3 (team) | 1 |
Links
Quick Verdict
- •burn-cuda leads with a health score of 74/100, but none of the options score above 80.
- •burn-cuda is depended on by 368 crates — strongest ecosystem trust.