# ScaNN

> ScaNN (Scalable Nearest Neighbors) is a pure ANN index library using anisotropic vector quantization and scorers for high-recall, high-throughput search at billion-scale. Features CPU/GPU support, TensorFlow/Numpy bindings, advanced quantization. For custom vector engines in recommendations, benchmarks; superior recall/throughput vs Faiss, building block unlike full Qdrant.

- **URL:** https://github.com/google-research/scann
- **Category:** ANN Indexing Libraries
- **Tags:** Pure ANN, Index Only, Benchmark Tool
- **Updated:** 2026-04-24 10:19
- **Canonical page:** https://vectordb.ever.works/items/scann

## Details

# ScaNN

ScaNN (Scalable Nearest Neighbors) is Google Research's library for efficient high-dimensional vector similarity search.

## Disk Persistence vs In-Memory

Optimized quantization reduces memory footprint, supporting disk-persistent deployments for datasets beyond RAM. Enables out-of-core search in large-scale setups unlike memory-intensive HNSW without compression.

## Key Features

- Asymmetric and anisotropic vector quantization
- High-throughput MIPS
- Up to 2x faster than benchmarks at same accuracy
- TensorFlow/Numpy support

## Use Cases

- Billion-scale embedding retrieval
- Production recommendation systems
- Compressed large vector collections

## Comparisons

**DiskANN vs HNSW**: ScaNN focuses on quantization for speed/accuracy trade-off, complementing graph indexes like DiskANN (graph+quantization) and outperforming uncompressed HNSW in compressed disk scenarios.

## Pricing

Open-source, free.

---

_This Markdown mirror is generated by Ever Works for AI agents. The canonical HTML page is at https://vectordb.ever.works/items/scann._