Decision Workspace
ternlang-compress vs bitnet-quantize vs tritium-core
Side-by-side comparison of Rust crates
51
ternlang-compress
experimentalv2.0.0
LLM-to-ternary compression pipeline — quantize float models to {-1,0,+1}, build sparse zero-index, export .tern files for ternlang-ml inference.
52
bitnet-quantize
experimentalv0.2.1
Microsoft BitNet b1.58 quantization and inference for Rust
57
tritium-core
experimentalv1.1.0-rc.2
Foundational ternary types, dtypes, scaling schemes, and reference math for Tritium.
Core Metrics
| ternlang-compress | bitnet-quantize | tritium-core | |
|---|---|---|---|
| Health Score | 51 | 52 | 57 |
| Total Downloads | 191 | 1.6K | 1.5K |
| 30d Downloads | 0 | 0 | 0 |
| Dependents | 0 | 12 | 38 |
| Releases | 11 | 3 | 2 |
| Last Updated | 51d ago | 254d ago | 35d ago |
| Age | 5m | 8m | 2m |
Health Breakdown
ternlang-compress
Maintenance
16
Quality
18
Community
6
Popularity
3
Documentation
8
bitnet-quantize
Maintenance
10
Quality
14
Community
9
Popularity
4
Documentation
15
tritium-core
Maintenance
12
Quality
19
Community
10
Popularity
4
Documentation
12
Technical Details
| ternlang-compress | bitnet-quantize | tritium-core | |
|---|---|---|---|
| Version | 2.0.0 | 0.2.1 | 1.1.0-rc.2 |
| Stable (≥1.0) | ✓ Yes | ✗ No | ✓ Yes |
| License | LicenseRef-BSL-1.1 | MIT | Apache-2.0 |
| Dependencies | 13 | 12 | 2 |
| Crate Size | 19KB | 58KB | 13KB |
| Features | 6 | 4 | 4 |
| Yanked % | 0.0% | 0.0% | 0.0% |
| Edition | 2024 | 2021 | 2024 |
| MSRV | — | 1.92 | 1.96 |
| Owners | 1 | 1 | 1 |
Links
Quick Verdict
- •tritium-core leads with a health score of 57/100, but none of the options score above 80.
- •bitnet-quantize has the most downloads (1.6K), suggesting wider adoption.
- •bitnet-quantize is pre-1.0 — API may change.