Full-text search benchmark

Search Benchmark, the Game: Tantivy's benchmark for full-text search engines.

corpus
5M Wikipedia docs · 962 queries
machine
GCP c3-highcpu-8
run
nightly of Nov 7, 2026

Latency vs Lucene 10.5, lower is faster

  1. 1 Infino 0.11.26 top-k: 0.46× counts: 0.38×
  2. 2 Infino main top-k: 0.48× counts: 0.38×
  3. 3 iresearch 26.10.3 top-k: 0.48× counts: 0.37×
  4. 4 Lucene 10.5.0, doc IDs reordered top-k: 0.78× counts: 0.85×
  5. 5 Lucene 10.5.0 top-k: 1.00× counts: 1.00×
  6. 6 Tantivy 0.27 top-k: 1.14× counts: 0.78×

Reproduce

# needs Cargo, and Java for Lucene; a full run takes over half an hour
git clone https://github.com/infino-ai/search-benchmark-game && cd search-benchmark-game
make corpus    # download and unpack Wikipedia
make index     # build each engine's index
make compile   # build the query runners
make bench     # run the queries, writes results.json
make serve     # dashboard at http://localhost:8080/

Questions

Is Infino faster than Lucene in this benchmark?

In the nightly run of Nov 7, 2026, released Infino 0.11.26 is 54% faster than Lucene 10.5 on top-k search and 62% faster on counts. iresearch is 0.01 faster than Infino on counts.

Can I compare these timings with the vector or ClickBench numbers?

No. Each benchmark here uses its own machines and method.

Which versions are measured?

The nightly run measures the released Infino crate and the engine at main. The other engines run at the versions pinned in the harness.

Full-text, vectors and SQL in one engine. Try it on your own data.

or run it yourself: pip install infino · Apache-2.0 · quickstart →