Embedding + vector query cost for a semantic search feature.
Semantic search queries are cheap; embedding and storing the corpus is where the money goes. Size those on the embeddings calculator and vector database calculator.
The Semantic Search Cost Calculator estimates your monthly bill for running semantic search, which combines two priced operations. For every query, the text must first be turned into a vector using an embedding model, then compared against your stored vectors in a vector database. The calculator multiplies your searches per month by the query tokens per search to get total tokens, prices those at your embedding rate per million tokens, and adds the vector query cost per million for the lookups themselves. The main drivers are simply how often you search and how many tokens each query carries.
The key trade-off to watch is query length: verbose queries multiply your embedding cost linearly, so trimming boilerplate or capping input tokens directly lowers the bill. Also note that embedding and vector-query pricing scale independently—a high search volume with short queries may be dominated by per-query lookup fees rather than tokens. Adjust each input to see which side dominates before committing to a provider or an indexing strategy.
Per query it's cheap: a small embedding of the query plus a vector database lookup. The bigger costs are embedding your entire corpus once (and re-embedding on updates) and storing the vectors.
Not the queries — they're fractions of a cent. It's indexing: embedding a large corpus and paying for vector storage. Budget those, and queries are nearly free.