Decision Workspace
turboquant-rs vs mistralrs-kv-cache vs tq-kv
Side-by-side comparison of Rust crates
49
turboquant-rs
experimentalv0.4.1
TurboQuant KV-Cache Quantization — 3-bit compression with zero accuracy loss (Zandieh et al., ICLR 2026)
52
mistralrs-kv-cache
experimentalv0.3.0
Trait interface for compressed KV-cache implementations in mistral.rs
49
tq-kv
experimentalv0.6.0
TurboQuant KV Cache Compression for LLMs — First working on GGUF Q4_K_M (3-Fix framework). KV Compaction (50x token reduction). Per-head adaptive bitwidth. 4-bit value compression. Calibration pipeline. SRHT QJL (115x faster). AVX2 SIMD fused attention. C FFI. 114 tests.
Core Metrics
| turboquant-rs | mistralrs-kv-cache | tq-kv | |
|---|---|---|---|
| Health Score | 49 | 52 | 49 |
| Total Downloads | 271 | 230 | 175 |
| 30d Downloads | 0 | 0 | 0 |
| Dependents | 0 | 4 | 0 |
| Releases | 5 | 3 | 6 |
| Last Updated | 36d ago | 36d ago | 55d ago |
| Age | 1m | 1m | 2m |
Health Breakdown
turboquant-rs
Maintenance
12
Quality
13
Community
6
Popularity
3
Documentation
15
mistralrs-kv-cache
Maintenance
12
Quality
14
Community
8
Popularity
3
Documentation
15
tq-kv
Maintenance
14
Quality
16
Community
6
Popularity
3
Documentation
10
Technical Details
| turboquant-rs | mistralrs-kv-cache | tq-kv | |
|---|---|---|---|
| Version | 0.4.1 | 0.3.0 | 0.6.0 |
| Stable (≥1.0) | ✗ No | ✗ No | ✗ No |
| License | MIT OR Apache-2.0 | MIT OR Apache-2.0 | MIT OR Apache-2.0 |
| Dependencies | 9 | 1 | 6 |
| Crate Size | 144KB | 17KB | 76KB |
| Features | 4 | 0 | 5 |
| Yanked % | 0.0% | 0.0% | 0.0% |
| Edition | 2021 | 2021 | 2021 |
| MSRV | 1.75 | 1.75 | — |
| Owners | 1 | 1 | 1 |
Links
Quick Verdict
- •mistralrs-kv-cache leads with a health score of 52/100, but none of the options score above 80.
- •turboquant-rs has the most downloads (271), suggesting wider adoption.