Annie Benchmarks & Performance

This section showcases Annie's performance and helps you optimize for your use case.

Interactive Benchmark Dashboard

If the dashboard does not load, view it here.

---

Library Comparison Table

Library	Build Time	Search Latency	Recall@10	Memory Usage	CPU	GPU Support
Annie	1x	1x	99.2%	1x	Yes	Yes
Faiss	1.2x	1.1x	98.7%	1.1x	Yes	Yes
Annoy	2.5x	2.2x	97.5%	1.3x	Yes	No
HNSWlib	1.1x	1.2x	98.9%	1.2x	Yes	No

All results normalized to Annie (lower is better for time/latency/memory).

---

Latency vs. Accuracy

• Annie achieves high recall with low latency compared to other libraries.

---

Memory Usage Benchmarks

• Annie is optimized for low memory usage, especially on large datasets.

---

Dataset Size Scaling

• Annie scales efficiently from 10K to 10M+ vectors.

---

GPU vs. CPU Performance

• GPU acceleration can provide 3-10x speedup for large batch queries.

---

Performance Tuning Recommendations

• Use batch operations for large queries.
• Tune index parameters (ef_search, ef_construction) for your workload.
• Monitor memory and CPU usage.
• Use GPU for large-scale or real-time workloads.

---

Explore Benchmarks

• Interactive Explorer
• Raw Results CSV

---

For more details, see Performance Optimization Tutorial.