HNSW stands for Hierarchical Navigable Small World. It is a graph index for approximate nearest neighbor search: each vector links to a handful of its nearest neighbors, and a query is answered by walking the graph toward the query point. The graph is built in layers, where a sparse top layer holds a thin sample of the data with long-range links and each layer down is denser, with the bottom layer containing every vector. Search finds the best node in a layer and uses it as the entry point for the layer below, which makes search cost scale logarithmically with the number of indexed vectors instead of linearly.

Why is HNSW approximate, and what is the tradeoff?

Greedy graph search stops when no neighbor is closer to the query, which is not guaranteed to be the true nearest neighbor. The size of the search candidate list, usually called ef or efSearch, controls how much of the graph is explored: widening it raises recall and raises query latency at the same time. Choosing that operating point is the actual engineering work. Mixpeek's vector store runs this under every query, so the knob exposed is recall rather than graph internals.

HNSW: How Vector Search Skips 10 Million Comparisons

Name: HNSW: How Vector Search Skips 10 Million Comparisons
Uploaded: 2026-07-28T20:00:00Z
Duration: 68 s

Multimodal Search in Action

Mixpeek Team

July 28, 2026

Summary

Ten million vectors sit in an index and a query arrives. Comparing against every one is exact and unusable, so vector databases run HNSW instead. This video is a live simulation of the algorithm: build a proximity graph, walk it greedily toward the query, then watch why a single flat graph makes that walk cross the entire space. HNSW stacks the graph into layers, so a thin top layer with long links covers enormous ground in a couple of hops and the best node found becomes the entry point one level down. Coarse to fine, logarithmic rather than linear. It is approximate, and widening the candidate list raises recall and latency together, which is the real operating decision. Full breakdown: mixpeek.com/guides/approximate-nearest-neighbor-algorithms

hnswvector-searchapproximate-nearest-neighborannembeddingsinformation-retrieval

HNSW: How Vector Search Skips 10 Million Comparisons

Summary

About this video

Frequently Asked Questions