Head-to-head,
on the same hardware.
Compare Infino's performance against the leading engines in three categories — search engines, vector databases, and SQL engines. Numbers coming soon.
Three tracks.
One Parquet format, three competitive surfaces. Each track has its own corpus, its own query mix, and its own reproducible harness.
vs the fastest search engines
BM25 search latency, ingestion throughput, and memory footprint on a Zipfian text corpus at 1M and 10M documents.
vs the fastest vector databases
ANN latency at recall@10 ∈ {0.90, 0.95, 0.99}, build throughput, and cold-open time on a 384-dimensional embedding corpus at 1M and 10M vectors.
vs the fastest SQL engines
Analytical query latency on the same Parquet bytes — predicate pushdown, GROUP BY, and joins on the same single copy of the data that BM25 and vector queries run against.
Read the architecture.
File layout, the inverted and vector index structures, and how the reader uses them.