--- title: Tuning Recall and Latency description: Trade recall against latency for vector search canonical: https://www.paradedb.com/docs/reference/vector/tuning --- Vector search is approximate: it probes a subset of the index's vector clusters rather than scanning every vector. Probing more clusters improves recall (i.e. how often the true nearest neighbors are returned) at the cost of higher latency. One session-level setting controls this tradeoff: ```sql SET paradedb.vector_cluster_max_probe = 100; ``` Controls the global probe budget in units of an average cluster's search cost, adjusted for filter selectivity. Accepts integers between `1` and `1000000`. A higher number increases recall at the cost of higher latency.