Benchmarks

Vector benchmark

#1 at 1M and 10M

VectorDBBench, Milvus's benchmark for vector databases

  1. 1 Infino QPS at recall ≥ 0.95: 7,741
  2. 2 OceanBase QPS at recall ≥ 0.95: 5,846
  3. 3 Elasticsearch QPS at recall ≥ 0.95: 5,115
  4. 4 Milvus QPS at recall ≥ 0.95: 4,974
  5. 5 pgvector QPS at recall ≥ 0.95: 2,301

Full-text search benchmark

54% faster than Lucene on top-k

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

  1. 1 Infino 0.11.26 top-k vs Lucene, lower is faster: 0.46×
  2. 2 Lucene, reordered top-k vs Lucene, lower is faster: 0.77×
  3. 3 Lucene top-k vs Lucene, lower is faster: 1.00×
  4. 4 Tantivy top-k vs Lucene, lower is faster: 1.13×

Analytical benchmark

#1 of 6 search engines

ClickBench, ClickHouse's benchmark for analytical workloads

  1. 1 Infino SQL score, lower is better: 1.55×
  2. 2 OpenSearch SQL score, lower is better: 16.40×
  3. 3 Quickwit SQL score, lower is better: 19.39×
  4. 4 Elasticsearch SQL score, lower is better: 28.52×
  5. 5 SigLens SQL score, lower is better: 90.11×
  6. 6 ParadeDB SQL score, lower is better: 95.80×

Embedded vector benchmark

0.963 recall@10 at 1.5 ms

retrievalbench, a benchmark for embedded vector search libraries

  1. 1 Infino flat_ivf recall@10: 0.963
  2. 2 FAISS PQ 8-bit recall@10: 0.950
  3. 3 turbovec 4-bit recall@10: 0.948
  4. 4 turbovec 2-bit recall@10: 0.835
  5. 5 FAISS PQ fastscan recall@10: 0.672

Accuracy benchmark

#1 in accuracy

OfficeQA Pro, Databricks' benchmark for grounded reasoning over documents

  1. 1 Supermodel + Claude Opus 5.5 accuracy: 73.3%
  2. 2 Mixedbread Toast + GPT-5.6 Sol accuracy: 70.0%
  3. 3 Databricks Genie + Claude Fable 5 accuracy: 60.0%
  4. 4 Snowflake Cortex Agent accuracy: 57.8%
  5. 5 Claude Opus 5.5 alone accuracy: 55.6%

Questions

Can I reproduce these numbers?

Yes for the four engine benchmarks. Each page has the command, and every published result names the engine version it ran. The OfficeQA page links Databricks' benchmark repository.

Where is Infino behind?

On ClickBench, Infino is in the top 15% of all systems for hot query speed on c8g.metal-48xl. Umbra, CedarDB, DuckDB and ClickHouse are ahead.

Measured. Now try it on your own data.

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