ANN-Benchmarks

ANN-Benchmarks is a benchmarking platform specifically for evaluating the performance of approximate nearest neighbor (ANN) search algorithms, which are foundational to vector database evaluation and comparison.

About this tool

ANN-Benchmarks

ANN-Benchmarks is a benchmarking environment for evaluating the performance of approximate nearest neighbor (ANN) search algorithms. It provides a platform to compare algorithms across multiple datasets and distance measures.

Features

  • Benchmarking Platform: Offers comprehensive benchmarking for a wide range of approximate nearest neighbor algorithms.
  • Multiple Datasets: Includes datasets such as glove, nytimes, fashion-mnist, gist, sift, word2bits, kosarak, and more, covering various data types and dimensionalities.
  • Variety of Distance Measures: Supports benchmarking on different distance metrics, including Angular, Euclidean, Hamming, and Jaccard.
  • Algorithm Coverage: Benchmarks a large selection of algorithms, including faiss-ivf, scann, pgvector, annoy, glass, hnswlib, BallTree(nmslib), vald(NGT-anng), hnsw(faiss), qdrant, n2, Milvus(Knowhere), mrpt, redisearch, SW-graph(nmslib), pynndescent, vearch, flann, luceneknn, weaviate, puffinn, elastiknn-l2lsh, sptag, ckdtree, opensearchknn, datasketch, and more.
  • Interactive Results: Provides interactive plots and detailed statistics for each benchmark, including recall, queries per second, index size, and build time.
  • Open Source Collaboration: Users can contribute new algorithms or improvements via pull requests on GitHub.
  • Results Visualization: Results are split by distance measure and dataset, with summary plots for quick comparison.

Pricing

No pricing information is provided; ANN-Benchmarks is an open-source benchmarking platform.

Links

Information

PublisherFox
PublishedMay 13, 2025