Home / Scale and cost
Scale and cost guides
7 published guides. Start with Scaling semantic search.
What changes when semantic search goes from thousands to billions of documents, and where it breaks first.
Read the guideLearning path
All guides
Understand it1 guides
How it works3 guides
ExplainerSearch latencyA latency budget for each stage from embedding the query to reranking, and how to cut each one.ExplainerCost of semantic searchEmbedding, storage, compute and reranking costs modelled per million documents and queries.ExplainerReal-time indexing and index freshnessHow to keep the index in step with changing data: streaming updates, deletes and re-indexing.
Compare and choose3 guides
Decision guideManaged vs self-hosted searchWhen a managed service pays for itself and when running your own vector search is cheaper and safer.Decision guideGPU acceleration for vector searchWhere GPUs speed up indexing and search, and the workloads where they do not pay off.Decision guideRecall vs latency trade-off in semantic searchHow index settings trade recall for speed, how to plot the curve for your data, and where to set the dial.