# Benchmarks — AxHash v0.11.0
Benchmarks were conducted on Apple Silicon M4, using the `release` profile. Results include:
- **Criterion.rs** for internal micro-benchmarks (per API, per payload).
## Latest Benchmark Recap (2026-05-29, Apple Silicon M4)
**Throughput (GB/s, higher is better):**
- 4B: 4.95 GiB/s
- 16B: 20.04 GiB/s
- 64B: 31.62 GiB/s
- 256B: 24.53 GiB/s
- 4K: 34.91 GiB/s
- 64K: 34.46 GiB/s
**Latency (ps/ns, lower is better):**
- 4B: ~752 ps
- 16B: ~743 ps
- 64B: ~1.88 ns
- 256B: ~9.72 ns
- 4K: ~109.3 ns
- 64K: ~1.77 µs
**Hash Trait Integration:**
- `axhash_of::<u64>()`: ~384 ps
- `axhash_of::<struct>()`: ~604 ps
- `axhash_of::<&str>()`: ~2.17 ns
**HashMap Workload (ops/s):**
- u64 Insert (100K): 109.4 Melem/s
- Get-hit (100K): 644 Melem/s
**Hash Quality:**
- Avalanche: 49.84%
- Collision: 0%
- Max bit-bias: 0.41%
**Note:**
- Outliers in measurements range from 1–19% (generally high and severe).
- Very stable performance on large payloads (≥4K).
- Very low latency on small payloads (≤16B).
---