Search engines and vector databases

In short: add a search engine when users need fast, typo-tolerant search, and skip a separate vector database at first: pgvector adds AI search to Postgres for free. When you do need a service, search engines bill per search or by server size, and vector databases bill by storage and queries. All prices below were checked in October 2026.

Two kinds of search

A search engine finds text that matches what someone typed, quickly, and forgives typos. Think of the search box on a shop or a documentation site. Each item you can search, such as a product or a post, is a record (or document).

A vector database searches by meaning. You turn each piece of text into an embedding: a list of numbers, made by an AI model, that captures what the text is about. Each embedding is stored as a vector. To search, you turn the question into a vector too and look for the closest ones. This is a common way for "chat with your documents" features to find the right paragraphs.

Search engines

Algolia

Algolia has a Free plan with 10,000 search requests a month and 50,000 records. Over quota, search is blocked until the next month or until you upgrade. Free applications with no searches, indexing or setting changes for 2 months are suspended and later deleted.

The Grow plan has no monthly fee. It includes 10,000 search requests and 100,000 records, then costs $0.50 per 1,000 extra requests and $0.40 per 1,000 extra records. Grow Plus adds AI features and costs $1.75 per 1,000 extra requests.

One catch: with search-as-you-type, every keystroke counts as a search request, and so do empty queries and filter changes. A real bill can be several times the number of searches people think they did.

Meilisearch Cloud

Meilisearch Cloud's Build plan is $30 a month and includes 50,000 searches and 100,000 documents. Extra searches cost $0.40 per 1,000 and extra documents $0.30 per 1,000. There is a 14-day free trial with no card, but no permanent free plan. Meilisearch is open source, so you can also host it yourself.

Typesense Cloud

Typesense Cloud rents you a dedicated cluster sized by memory, with no limit on records or searches. The smallest has 0.5 GB of memory and costs $0.03 an hour, or $21.60 a month. The first 720 hours of a cluster are free. Memory is what limits you, so the size you need depends on how much data you index. For production, Typesense recommends High Availability, which runs 3 or more nodes and multiplies the price. Typesense is also open source.

How they compare

The stack builder assumes 20 searches per user a month and 10 searchable records per user.

Monthly users	Searches	Algolia	Meilisearch	Typesense
1,000	20,000	$5.00	$30.00	$21.60
10,000	200,000	$95.00	$90.00	$21.60
100,000	2,000,000	$1,355	$1,080	$21.60

Typesense stays flat only while your data fits in 0.5 GB of memory. Above that, you need a bigger cluster. Algolia and Meilisearch grow with every search.

Vector databases

pgvector

pgvector is a free, open-source Postgres extension. It stores vectors next to your normal data, inside the Postgres database you already pay for, such as Supabase or Neon. Many hosted Postgres providers include it.

Each vector takes 4 bytes per dimension plus 8 bytes, so a common 1,536-dimension vector is about 6 KB. Indexes can cover up to 2,000 dimensions, or 4,000 with half-precision vectors. Vectors use your database's storage, so check your plan's limits.

Pinecone

Pinecone's free Starter plan includes 2 GB of storage and 1 million read units a month. Past a limit, requests fail until you upgrade. Starter indexes run only on AWS us-east-1.

Builder is a flat $20 a month for 10 GB and 2 million read units. Usage past those limits is blocked, not billed. Standard has a $50 monthly minimum, with storage at $0.33 per GB a month and read units from $16 per million.

A query uses at least 0.25 read units, and 1 read unit per GB of the namespace (a group of vectors) it searches. One small namespace per user keeps queries at the minimum.

Qdrant Cloud

Qdrant Cloud includes one free cluster with 0.5 vCPU, 1 GB of memory and 4 GB of disk. Qdrant says it holds about 1 million vectors of 768 dimensions. A free cluster unused for a week is suspended, and one inactive for 4 weeks is deleted. Qdrant does not publish fixed prices for paid clusters, only an in-app calculator, so the builder lists the free cluster only.

turbopuffer

turbopuffer has no free tier. It has a $16 monthly minimum, which covers your first $16 of storage (up to $0.33 per GB a month) and queries. Months with less usage still cost $16. The minimum was $64 a month until June 2026, so older articles quote the higher price.

How much you store

The builder assumes 100 vectors per user, at 6 KB each. That is 0.6 GB for 1,000 users and 6 GB for 10,000. At 1,000 users, Pinecone Starter and Qdrant's free cluster both fit. At 10,000 users, 6 GB is over Pinecone Starter's 2 GB, so the builder picks Builder at $20 a month, and it is over the 4 GB disk on Qdrant's free cluster.

What to pick

• Site or product search, small scale: Algolia's free plan, or Grow if you pass 10,000 searches. Watch search-as-you-type.
• Growing search traffic: Typesense Cloud's flat cluster price, as long as your data fits in memory. Meilisearch if you want to self-host later.
• AI search on Postgres: pgvector. No new service and no new bill.
• A separate vector database: Pinecone Starter or Qdrant's free cluster while you test, then Pinecone Builder for a predictable $20.

Compare them in the stack builder, or browse Search Engines and Vector Databases.