Decision Workspace
ramvamp vs frink-inference vs ferrox-inference
Side-by-side comparison of Rust crates
40
ramvamp
experimentalv0.1.0
Run 26-30B fine-grained MoE models in about 3 GB of RAM by streaming experts from NVMe. No GPU required.
52
frink-inference
experimentalv0.49.0
Pure-Rust GGUF / MoE inference engine: mmap loaders, quantized CPU + Metal + CUDA kernels
47
ferrox-inference
experimentalv0.25.0
Pure-Rust GGUF / MoE inference engine: mmap loaders, quantized CPU + Metal + CUDA kernels
Core Metrics
| ramvamp | frink-inference | ferrox-inference | |
|---|---|---|---|
| Health Score | 40 | 52 | 47 |
| Total Downloads | 18 | 324 | 353 |
| 30d Downloads | 0 | 0 | 0 |
| Dependents | 0 | 0 | 0 |
| Releases | 1 | 24 | 23 |
| Last Updated | 61d ago | 16d ago | 20d ago |
| Age | 2m | 20d | 1m |
Health Breakdown
ramvamp
Maintenance
9
Quality
11
Community
6
Popularity
2
Documentation
12
frink-inference
Maintenance
18
Quality
15
Community
6
Popularity
3
Documentation
10
ferrox-inference
Maintenance
13
Quality
15
Community
6
Popularity
3
Documentation
10
Technical Details
| ramvamp | frink-inference | ferrox-inference | |
|---|---|---|---|
| Version | 0.1.0 | 0.49.0 | 0.25.0 |
| Stable (≥1.0) | ✗ No | ✗ No | ✗ No |
| License | MIT OR Apache-2.0 | Apache-2.0 | Apache-2.0 |
| Dependencies | 8 | 7 | 7 |
| Crate Size | 146KB | 11KB | 11KB |
| Features | 0 | 4 | 4 |
| Yanked % | 0.0% | 0.0% | 0.0% |
| Edition | 2024 | 2021 | 2021 |
| MSRV | 1.88 | — | — |
| Owners | 1 | 1 | 1 |
Links
Quick Verdict
- •frink-inference leads with a health score of 52/100, but none of the options score above 80.
- •ferrox-inference has the most downloads (353), suggesting wider adoption.