Decision Workspace
bitnet-inference vs tritium-core vs drama_llama
Side-by-side comparison of Rust crates
47
bitnet-inference
growingv1.0.0
High-performance inference engine for BitNet models
57
tritium-core
experimentalv1.1.0-rc.2
Foundational ternary types, dtypes, scaling schemes, and reference math for Tritium.
54
drama_llama
growingv0.9.0
Local LLM inference behind an Anthropic-shaped API: llama.cpp and Metal-MoE backends, chat sessions with prefix caching, per-model tool-call dialects, and GBNF-constrained structured output.
Core Metrics
| bitnet-inference | tritium-core | drama_llama | |
|---|---|---|---|
| Health Score | 47 | 57 | 54 |
| Total Downloads | 1.5K | 1.5K | 33.1K |
| 30d Downloads | 0 | 0 | 0 |
| Dependents | 0 | 38 | 5 |
| Releases | 4 | 2 | 28 |
| Last Updated | 405d ago | 35d ago | 4d ago |
| Age | 1y 2m | 2m | 2y 5m |
Health Breakdown
bitnet-inference
Maintenance
7
Quality
18
Community
5
Popularity
4
Documentation
13
tritium-core
Maintenance
12
Quality
19
Community
10
Popularity
4
Documentation
12
drama_llama
Maintenance
19
Quality
12
Community
8
Popularity
5
Documentation
10
Technical Details
| bitnet-inference | tritium-core | drama_llama | |
|---|---|---|---|
| Version | 1.0.0 | 1.1.0-rc.2 | 0.9.0 |
| Stable (≥1.0) | ✓ Yes | ✓ Yes | ✗ No |
| License | MIT OR Apache-2.0 | Apache-2.0 | non-standard |
| Dependencies | 32 | 2 | 54 |
| Crate Size | 158KB | 13KB | 3.5MB |
| Features | 12 | 4 | 20 |
| Yanked % | 0.0% | 0.0% | 0.0% |
| Edition | 2021 | 2024 | 2021 |
| MSRV | — | 1.96 | 1.97 |
| 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.
- •drama_llama has the most downloads (33.1K), suggesting wider adoption.
- •⚠ bitnet-inference has not been updated in over a year.
- •drama_llama is pre-1.0 — API may change.