Search pricing compared
Some search vendors bill per request and record; others bill for cluster size. Test relevance and peak load before switching.
11 services · sources last checked 2026-10-06 · how we compare
What the most-compared services cost
Estimated yearly bill at 1M records, 5M searches / month. Enter your own usage below.
- Algolia$34,260/yrGrow · $360 records + $2,495 searches · Published rates
- Meilisearch Cloud$27,360/yrUsage-based · $270 records + $1,980 searches · Modeled rates
- Elastic Cloud (Elasticsearch)$2,520/yrServerless, production (about 20 GB data) · $0 records + $0 searches · Modeled rates
- Typesense Cloud$864/yr4 GB RAM, 2 vCPU burst 4 h/day · $0 records + $0 searches · Modeled rates
Enter your usage to estimate each bill. Services differ in what they include; check the rates, assumptions and extra charges below.
The estimates below use the default usage. Choose your current provider to compare costs.
| Provider | Parent jurisdiction | Features / pricing source | Est. monthly | Est. yearly |
|---|---|---|---|---|
| BonsaiManaged Elasticsearch and OpenSearch on AWS and GCP with flat monthly plans capped by document count, storage and concurrency. | 🇺🇸 United States | EU residencyGDPR DPAModeled ratesAssumes roughly 1 KB per record and checks document counts, not concurrency. Index overhead, peak traffic and availability can require a larger plan. | $50Standard Elm · $0 records + $0 searches | $600 |
| Typesense CloudManaged clusters of the open-source Typesense engine, billed hourly by RAM and vCPU with no per-search or per-record fees. | 🇺🇸 United States | EU residencyGDPR DPAopen sourceModeled ratesSingle-node cluster with 1 KB records and modeled sizing. Production HA roughly triples cost; EU regions can cost more. | $724 GB RAM, 2 vCPU burst 4 h/day · $0 records + $0 searches | $864 |
| Amazon OpenSearch ServiceAWS-managed OpenSearch (open-source Elasticsearch fork). Provisioned domains bill per instance-hour plus EBS storage; Serverless bills per OCU-hour. | 🇺🇸 United States | EU residencyGDPR DPAopen sourceModeled ratesSingle-node domain sized at 1 KB per record. Multi-zone availability, replicas and peak query load are not modeled. | $891x m6g.large.search + 50 GB gp3 · $0 records + $0 searches | $1,063 |
| Elastic Cloud (Elasticsearch)Managed Elasticsearch from Elastic. Serverless projects bill per VCU-hour of search and ingest compute plus GB stored; hosted clusters bill by size. | 🇳🇱 Netherlands | EU residencyGDPR DPAopen sourceModeled ratesUses published example deployments and assumes 1 KB records. Real query load and resource sizing can differ. | $210Serverless, production (about 20 GB data) · $0 records + $0 searches | $2,520 |
| Vespa CloudManaged Vespa, the open-source big-data search and ranking engine spun out of Yahoo. Billed per vCPU, memory and disk hour. | 🇳🇴 Norway | EU residencyGDPR DPAopen sourcePublished rates | $695Basic, minimal production (2 container + 2 content nodes, 8 GB) · $0 records + $0 searches | $8,340 |
| Meilisearch CloudManaged Meilisearch from Paris-based Meili. Choose usage-based billing (per document and search) or resource-based instances. | 🇫🇷 France | EU residencyGDPR DPAopen sourceModeled ratesResource sizing assumes 1 KB records. Large usage-based estimates may require a negotiated plan. | $2,280Usage-based · $270 records + $1,980 searches | $27,360 |
| AlgoliaHosted search-as-a-service billed per search request and per stored record. Pay-as-you-go Grow plan; enterprise features need an Elevate contract. | 🇺🇸 United States | EU residencyGDPR DPAPublished rates | $2,855Grow · $360 records + $2,495 searches | $34,260 |
| CoveoEnterprise AI search and relevance platform for commerce, service and workplace, sold by annual contract priced per query volume. | 🇨🇦 Canada | GDPR DPAcontractNo listed rates | —Contract only · no public price | — |
Resource-priced vendors (billed by RAM/cluster) are mapped to the smallest plan that fits; see notes.
Need EU residency? Turn on the EU filter to pick a plan that includes it, which may cost more. “EU plan available” alone does not mean the price shown includes residency. Shared links save your inputs and filters, but prices can change.
Looking for a European-owned provider? See European search providers.
Self-hosted options
Operating cost not estimated · servers, licenses and engineering time are yours. We list these separately because there’s no calculated bill to compare with managed services.
- Meilisearch (self-hosted) — Open-source (MIT community edition) search engine in Rust with typo tolerance, filters and hybrid vector search, run on your own servers.
- Orama (OramaJS) — Open-source (Apache-2.0) full-text, vector and hybrid search library in TypeScript that runs in the browser, at the edge or on your server.
- Typesense (self-hosted) — Open-source (GPL-3.0) in-memory typo-tolerant search engine written in C++, run on your own servers.
No self-hosted options match these filters.
Head-to-head comparisons
Search pricing explained
What drives the bill
Search vendors use two pricing models. Usage-based vendors charge per search request and per record stored, so the bill grows with traffic. Resource-based vendors charge for the RAM and CPU of a cluster, so the bill stays flat until you need a bigger cluster. Open-source engines such as Typesense and Meilisearch can also be self-hosted on one modest server for many workloads.
Costs the calculator leaves out
- Search-as-you-type. Every keystroke can be a billable request unless you debounce it.
- Replicas and sort variants. On some vendors every alternative sort order duplicates your records.
- AI and semantic search features (vector search, re-ranking, recommendations), which are often priced separately.
- High availability. Some engines recommend three-node clusters, which cost more than the single-node examples.
When paying more makes sense
Merchandising tools, analytics and A/B testing let non-engineers tune results, which can be valuable for e-commerce. If you mainly need text search and filters, try a resource-priced or self-hosted engine. Compare results for your most common queries, including typos and queries that return nothing.