Every European vector database reviewed
Berlin, Germany
Free tier forever (single-node cluster, 0.5 vCPU, 1 GB RAM, 4 GB disk) / Standard usage-based, 99.5% uptime SLA / Premium requires a minimum spend, 99.9% uptime SLA
Free tier, forever
Best for: Teams building semantic search or RAG on a German vendor
Qdrant is a purpose-built vector database, open source, that a team can self-host or run on Qdrant Cloud. The pricing page lists a free tier of a single-node cluster with 0.5 vCPU, 1 GB RAM and 4 GB disk that is free forever, a usage-based Standard tier for production with dedicated resources and a 99.5% uptime SLA, and a Premium tier with a minimum spend, SSO, private VPC links and a 99.9% SLA.
The legal picture is clear. Qdrant Solutions GmbH publishes an impressum with its address at Chausseestrasse 86 in Berlin, the register entry HRB 235335 B at the Berlin-Charlottenburg court and André Zayarni as representative. The terms name Qdrant Solutions GmbH as the company and apply the law of Germany. A single-purpose engine is the point: it is narrower than a general-purpose database, and the managed cloud, rather than a broad platform, is what you buy.
What Qdrant does well
- German GmbH with a full imprint and commercial-register number
- Free single-node tier that stays free
- Open source, so self-hosting on European servers is possible
- Dedicated purpose-built vector engine
- Terms governed by the law of Germany
Where Qdrant falls short
- Single-purpose: not a general-purpose database
- Standard and Premium prices are usage-based or on request
- Region list is not on the pricing page
- The free tier is small at 4 GB disk
Standout feature. A German, open-source vector engine with a permanent free tier and a published register entry.
Amsterdam, Netherlands
Free tier $0/month (100,000 objects, 1 cluster per user) / Flex from $45/month pay-as-you-go / Premium prepaid contract, contact sales
Free tier, no credit card required
Best for: Teams wanting hybrid search from a Dutch vendor
Weaviate is the second purpose-built vector database here, and its pricing page lists a free tier at $0 a month with 100,000 objects, 1 GB memory and one cluster per user, a Flex plan from $45 a month on pay-as-you-go, and a Premium prepaid contract on request. Embeddings and a Query Agent are offered as services inside Weaviate Cloud, and the plans mention coverage on AWS, GCP and Azure.
The privacy policy is issued by Weaviate B.V., a private limited company at Prinsengracht 769A in Amsterdam, and names the Dutch Data Protection Authority. Compared with Qdrant, the cloud runs on the three US hyperscalers, which is a point to check against your own requirements, and the pricing page does not list the regions. For teams that want a Dutch contracting party and a hybrid-search engine, it is a sound choice.
What Weaviate does well
- Dutch B.V. named in the privacy policy
- Free tier with 100,000 objects
- Flex plan from $45 a month with no commitment
- Open-source core
- Built-in embeddings and query services
Where Weaviate falls short
- Cloud runs on AWS, GCP and Azure
- Regions are not listed on the pricing page
- Premium is quoted on request
- Free tier allows one collection and three tenants
Standout feature. A Dutch-registered, open-source vector database with hybrid search and a $45 entry plan.
Trondheim, Norway
Vespa Cloud pricing is not shown on the pricing page; free trial and contact form only
Free trial
Best for: Large-scale search and ranking beyond plain vector similarity
Vespa describes itself as letting you query, organise and make inferences in vectors, tensors, text and structured data, scaling to billions of items and thousands of queries per second. Its vector-database page stresses hybrid ranking over lexical, semantic and structured signals in one pipeline, and multi-vector representations. It is the heaviest engine in this directory, aimed at teams whose search quality depends on custom ranking.
The privacy policy names Vespa.ai AS as data controller for its websites and for Vespa Cloud contracts, and the Norwegian Bronnoysund register lists VESPA.AI AS in Trondheim, organisation number 931605569. Norway is inside the EEA but not the EU. The pricing page shows no amounts, only a free trial and a contact form, and the licence was not stated on the pages checked.
What Vespa does well
- Norwegian company in the official register
- Hybrid ranking and multi-vector support in one engine
- Scales to billions of documents by its own account
- Free trial of Vespa Cloud
Where Vespa falls short
- No prices published
- More complex than a simple vector store
- Norway is in the EEA but not the EU
- Licence not stated on the pages checked
Standout feature. A Norwegian engine for teams where ranking, not just vector similarity, decides search quality.
Amsterdam, Netherlands
Elastic Cloud Serverless and Hosted, free trial; amounts are not in the server-rendered pricing page
Free trial
Best for: Teams already running Elasticsearch who want vectors in the same engine
Elastic markets Elasticsearch as a vector database with hybrid search, metadata filters in a single call and Better Binary Quantization claiming up to 95% memory savings. Since Elastic acquired Jina AI in October 2025, Jina models are offered for embeddings and reranking inside Elastic. Deployment options are Elastic Cloud Serverless, Elastic Cloud Hosted and self-managed Elasticsearch.
Elastic N.V. gives an Amsterdam address (Keizersgracht 281) in its privacy statement and says the statement covers the company and its subsidiaries. That makes the parent Dutch, but it is a large company with offices worldwide, and the contracting entity in an order form depends on the customer, which was not verified here. The server-rendered pricing page carries no amounts.
What Elasticsearch (Elastic) does well
- Dutch parent company with an Amsterdam address
- Familiar engine for existing Elasticsearch users
- Hybrid search and quantization built in
- Serverless, hosted and self-managed options
Where Elasticsearch (Elastic) falls short
- Large company with operations well beyond Europe
- Contracting entity per customer not verified
- No amounts on the pricing page text
- Broader and heavier than a pure vector store
Standout feature. The incumbent search engine from a Dutch parent company, now with vectors and hybrid search built in.
Paris, France
Cloud from $20/month (14-day trial) / usage-based plan $30/month base (100K docs, 50K searches) / resource-based from $23/month / Enterprise on request
14-day free trial, no credit card required
Best for: Product search that blends keywords and meaning
Meilisearch began as a typo-tolerant full-text engine and now lists vector and hybrid search, federated search and conversational AI among its features. It is open source and also sold as Meilisearch Cloud, with a 14-day trial, a usage-based plan at $30 a month for 100,000 documents and 50,000 searches, a resource-based XS instance at $23 a month, and an Enterprise plan on request.
The terms say Meilisearch is developed by Meili, a simplified joint-stock company under French law, registered in Paris under number 844 156 364, and that French law and the Paris courts apply. It is a better fit for site and app search than for a very large embedding store, which is where Qdrant, Weaviate and Vespa are stronger.
What Meilisearch does well
- French company, French law and Paris courts in the terms
- Hybrid search combining keywords and vectors
- Open source and self-hostable
- Prices published, from $20 a month
Where Meilisearch falls short
- Built for search products, not very large vector stores
- Cloud regions are not on the pricing page text
- Enterprise plan quoted on request
- Pricing is in US dollars
Standout feature. A French open-source search engine that adds vectors to typo-tolerant search with prices published from $20 a month.
London, United Kingdom
Cloud: first instance $0/hr with 1 GB storage free / further instances from $0.021/hr, compute and storage priced separately / Enterprise on request
Free instance
Best for: Teams wanting vectors, graph and documents in one database
SurrealDB combines graph, vector, document, relational and full-text search in one engine and one query language, and positions its newer Agent Memory layer for AI agents. The Cloud pricing page lists a first instance at $0 an hour with 1 GB of storage free, further instances from $0.021 an hour, compute and storage priced separately, and an Enterprise plan.
The site footer states that SurrealDB Ltd is registered in England and Wales, company number 13615201, with a registered address in Altrincham and a trading address in London. The United Kingdom is outside the EU and the EEA, though it holds an EU adequacy decision. It is a multi-model database more than a dedicated vector store, so it suits applications that need several data models together.
What SurrealDB does well
- Vectors, graph and documents in one engine
- Free first Cloud instance
- Hourly pricing is published
- Company number and addresses stated on the site
Where SurrealDB falls short
- UK entity, outside the EU and EEA
- Not a dedicated vector database
- Licence not stated on the pages checked
- Data location not stated on the pages checked
Standout feature. A British multi-model database that puts graph, document and vector search behind one query language.
Chatham, Kent, United Kingdom (registered office)
Open-source software; no price published on the pages checked
Best for: Self-hosted full-text plus vector search
Manticore Search is an open-source search database that its home page describes as covering full-text, typo-tolerant, vector, hybrid and conversational search, queried over SQL or HTTP on your own servers. A sample query on the page uses a knn function over an embedding column, and the vendor claims large speed advantages over Elasticsearch in its own benchmarks.
The privacy policy names Manticore Software Ltd as the operator, and Companies House lists MANTICORE SOFTWARE LTD, number 10772872, incorporated 16 May 2017 with a registered office in Chatham, Kent. It is a British company, outside the EU and EEA. No prices are published on the pages checked, so the draw is the open-source engine rather than a managed service.
What Manticore Search does well
- Open source and self-hosted on your own servers
- Full-text, vector and hybrid search in one engine
- SQL and HTTP interfaces
- Company confirmed in Companies House
Where Manticore Search falls short
- UK entity, outside the EU and EEA
- No published prices or managed-cloud details checked
- Smaller ecosystem than Elasticsearch
- Benchmarks are the vendor's own
Standout feature. A British open-source search database that runs full-text and vector queries over plain SQL on your own servers.
Helsinki, Finland
PostgreSQL and OpenSearch: Free $0/month / Developer $5/month / Hobbyist from $12/month / Startup from $75/month
Free plan, plus a free 30-day trial of larger plans
Best for: Managed PostgreSQL with pgvector or OpenSearch k-NN, EU-based
Aiven is not a vector database in itself. It runs managed open-source data services, and two of them cover vectors: PostgreSQL lists the pgvector extension among its available extensions, and OpenSearch lists k-NN among its plugins. The pricing page shows a free plan, a Developer plan at $5 a month, Hobbyist from $12 and Startup from $75, plus a free 30-day trial of larger plans.
Aiven Oy is in the Finnish business register (business ID 2795743-5), and the terms are governed by Finnish law with arbitration in Helsinki. For a team that already uses PostgreSQL, adding pgvector at a European provider keeps the stack simple. The cost is that it is not tuned for very large vector workloads the way Qdrant or Vespa are.
What Aiven does well
- Finnish company in the official register
- Free plan and $5 Developer plan published
- pgvector in PostgreSQL and k-NN in OpenSearch
- Terms under Finnish law
Where Aiven falls short
- A general platform, not a dedicated vector engine
- Free plans cannot choose a region
- Multi-cloud, so the underlying hyperscaler matters
- Prices are in US dollars
Standout feature. The simplest route to pgvector or OpenSearch k-NN from a Finnish company with published plans from $0.
Roubaix, France
Managed PostgreSQL billed per hour by plan; no vector-specific price on the page
Best for: pgvector on a French cloud, with Qdrant as a partner option
OVHcloud's PostgreSQL page names pgvector and pgvectorscale for storing, indexing and searching vector representations for AI, and also mentions Qdrant as a partner solution. This is vector search as a feature of managed PostgreSQL, not a separate product, so it suits teams that already run on OVHcloud.
The Public Cloud terms refer to OVH Groupe SA affiliates (537 407 926 RCS Lille Metropole, 2 rue Kellermann, 59100 Roubaix), and the page footer names OVH SAS. The company is French, and no vector-specific price is given on the page. Compared with Aiven, OVHcloud is an infrastructure provider first, with a wider catalogue and a heavier console.
What OVHcloud does well
- French company with a published register number
- pgvector and pgvectorscale named on the page
- Qdrant offered as a partner solution
- Part of a full European cloud catalogue
Where OVHcloud falls short
- No vector-specific price on the page
- Not a dedicated vector database
- Several OVH group entities appear in the terms
- Wider console than a single-purpose service
Standout feature. A French infrastructure provider that offers pgvector and a Qdrant partner option next to its other cloud services.
Paris, France
Managed PostgreSQL billed by usage; no vector-specific price on the page
Best for: Turning a French managed PostgreSQL into a vector database
Scaleway's database page says you can turn your database into a vector database with pgvector, alongside PostgreSQL extensions such as TimescaleDB and PostGIS, and that you pay for what you use with no up-front commitment. As with Aiven and OVHcloud, this is vector search inside a general managed database.
The page footer names Scaleway SAS. The company belongs to the Iliad Group, based in Paris. No vector-specific price is quoted on the page. For a small RAG feature in an application already on Scaleway it is the lowest-friction option; for a large embedding workload a purpose-built engine fits better.
What Scaleway does well
- French company, footer names Scaleway SAS
- pgvector offered on managed databases
- Usage-based, no up-front commitment
- Fits teams already on Scaleway
Where Scaleway falls short
- Not a dedicated vector database
- No vector-specific price on the page
- Register number not checked on the pages read
- Less specialised than Qdrant or Weaviate
Standout feature. The lowest-friction pgvector option for teams already on a French cloud.
Lausanne, Switzerland
Managed databases priced per hour in EUR, CHF or USD on the pricing page; pgvector is an extension of managed PostgreSQL, no separate price
Best for: pgvector under Swiss law with European data centres
Exoscale's product menu lists Vector Databases under its AI storage section, with pgvector described as vector similarity search inside PostgreSQL and Vector Search as semantic search over your own data. Managed databases are priced per hour on the pricing page in EUR, CHF or USD, and pgvector is an extension, not a separate product.
The terms name Akenes SA, Boulevard de Grancy 19A, Lausanne, as the supplier, apply the laws of Switzerland with the courts of the canton of Vaud, and the footer gives the registration CHE-423.524.322 and states that Exoscale is a member of A1 Digital. Switzerland is outside the EU and EEA, though it holds an EU adequacy decision.
What Exoscale does well
- Swiss supplier named with a registration number
- pgvector and vector search listed as products
- Prices in EUR, CHF or USD
- Terms under Swiss law
Where Exoscale falls short
- Switzerland is outside the EU and EEA
- Not a dedicated vector database
- Part of A1 Digital, a larger group
- No vector-specific price
Standout feature. A Swiss managed-PostgreSQL provider that lists vector databases as a product and prices in francs or euros.
