Fast enough to ask anything.
A graph is only useful if you can query it at the speed of a question. Most Whisper queries finish in under a millisecond of server time — at billions of nodes and edges, in one place.
Depth without the wait
Speed at this scale isn’t about a bigger machine; it’s about doing less work per question. Lookups land directly on an index, so anchoring a query on a name, an address, or a network is effectively instant no matter how large the graph gets.
From there, the traversals that matter are short because the graph is pre-joined — the cross-layer step that would be a deep scan elsewhere is already a single indexed edge here. And every query runs inside a bounded depth, so the cost of a question stays predictable instead of fanning out across billions of relationships. The result is the same whether the graph holds a million edges or forty billion: an answer at the speed of the next question.
Benchmarks, in detail
We’re putting together a fuller performance write-up — latency distributions, throughput under concurrency, and how query shape affects cost. It’ll land here. In the meantime, the fastest read on performance is to run something yourself: try a live query →