Latency you can reproduce
Concurrency · TPC-H
Flat latency under concurrent sessions
~2.7s
Whether 1 or 50 concurrent sessions
19 of 22 TPC-H queries (Q18, Q20, and Q21 excluded), sum of each query's best-of-three median execution time, result caching disabled, 1–50 concurrent streams. Read the full benchmark →
Vector & semantic search
Sub-100ms vector search, with filters
<100ms
Filtered vector search, predicates applied
1.2M rows, 768-dimensional embeddings (Sphere dataset). Server-side execution time, three warm runs after a discarded cold run. Read the full benchmark →