Open-source vector database with a managed cloud on AWS, GCP and Azure
Choose Weaviate if you build retrieval for RAG or agents and want an open core you can leave for self-hosting. G2 shows 52 user reviews of it. Skip the free cluster for anything near production, since it has no replication or backups and stops at 100,000 objects, and compare Pinecone and Qdrant before you commit.
Hosted embeddings and a Query Agent sit beside hybrid search inside one managed database.
An open BSD core lowers the cost of trying Weaviate and of leaving it for self-hosting, so lock-in rests on the cloud services layered on top rather than on the data format.
Main trade-off
The free cluster has no replication or backups, so a team must pay before it can test production behaviour.
| Plan | Price |
|---|---|
| Free (shared)Cloud · Learning and prototypes · AWS (2 regions) | $0/mo |
| Flex (shared)Cloud · Pilots and small production workloads · GCP and AWS (7 regions) | $45/mo |
| Premium (shared)Cloud · Production teams that want predictable spend · GCP and AWS (7 regions) | $400/mo |
| Premium (dedicated)Cloud · Regulated and large-scale production workloads · AWS, GCP and Azure (about 40 regions) | — |
The free cluster needs no card. Flex starts from a $45 monthly floor. Premium runs on a prepaid contract, and dedicated deployment is quoted by sales. Self-hosting the open-source database costs only your own servers.
Weaviate lists transfer as free during a limited promotional period and reserves the right to charge later with notice.
As stated by the provider. Ask for the current report before you rely on one.
Weaviate is a BSD-3-Clause vector database built for hybrid vector and keyword search over objects and their embeddings. Teams run it themselves or use the managed cloud offered across AWS, GCP and Azure, with hosted embeddings, a Query Agent and an independently audited SOC 2 Type II program.
Hybrid search
Combines vector similarity with keyword matching and structured filters in one query, and it is on every plan including Free.
Weaviate Embeddings
Hosts embedding models inside Weaviate Cloud, billed by usage, with 2,000 requests a day on the Free plan.
Query Agent
Turns natural-language questions into database operations, with 1,000 requests a month free and 30,000 a month on Flex.
Multi-tenancy and RBAC
Isolates tenants and controls access by role on every plan; the Free plan allows up to 3 tenants.
Vector compression and index choice
Offers compression on all plans and a choice of index types such as HNSW and HFresh on paid plans, trading memory against search quality.
Deployment choice
Runs as self-hosted open source, shared cloud, dedicated cloud on AWS, GCP or Azure, or a Bring Your Own Cloud setup.
Engram
A managed memory service that gives LLM agents persistent memory, listed in Weaviate's documentation.
| Service | Regions | Pricing | Compliance | Checked |
|---|---|---|---|---|
| Global | Free tier · Pay as you go · Monthly | SOC 2 · HIPAA | 16 Nov | |
| US · Europe · Asia Pacific | Free tier · Pay as you go · Subscription | SOC 2 · GDPR · ISO 27001 +1 | 13 Nov | |
| -- | Free tier · Pay as you go · Hourly | SOC 2 · HIPAA · GDPR | 15 Nov |
Nothing for the free cluster. Flex meters vector dimensions (from $0.00465 per million), storage (from $0.12 per GiB) and backups (from $0.03036 per GiB) above a $45 monthly floor. Premium is prepaid with a floor from $400 a month, and dedicated plans are priced by sales. Rates shift with cloud, region and index type.
Yes. Each user gets one free cluster with 100,000 objects, 1 GB of memory and 10 GB of disk, with one collection and up to 3 tenants. It has no replication and no backups. You can upgrade to Flex at any time.
Most of the repository is under the BSD 3-Clause license, and you can run the database yourself. Files in the wl/ folder are commercial and need a license key, and the Enterprise Edition features require one. Weaviate Cloud wraps the same project in a managed service.
The free cluster runs on AWS in 2 regions, and Flex and Premium shared run on GCP and AWS in 7 regions. Premium dedicated covers AWS, GCP and Azure across about 40 regions. Weaviate also lists a Bring Your Own Cloud option and a DigitalOcean managed offering.
Weaviate Cloud is SOC 2 Type II audited through Drata, and the report is available in its Trust Portal. Enterprise Cloud on AWS is HIPAA compliant, and the pricing table lists HIPAA only on the dedicated plan. The free and Flex plans do not include HIPAA.
Weaviate is an open-source vector database from Weaviate B.V., released under the BSD-3-Clause license, that stores objects together with their embeddings and answers queries that mix vector similarity, keyword matching and filters. The company marked its fourth birthday in July 2023, which puts the start in 2019, and its GitHub repository shows about 17,000 stars. You can run the database yourself or buy it as Weaviate Cloud on AWS, Google Cloud and Azure.
Its pricing page describes services that run inside the cloud product: Weaviate Embeddings for hosted embedding models, a Query Agent that turns natural-language questions into database operations, and, per its documentation, Engram, a memory service for agents. Hybrid search, vector compression, multi-tenancy and role-based access control come with each cloud plan. Read the Hokai page for Pinecone first when you shortlist, then the one for Qdrant.
It suits developers who build retrieval for RAG systems and agents, and G2 lists LangChain and LlamaIndex among its integrations. For a wider view of the category, read how Hokai weighs the retrieval stack in its ranking of search and retrieval tooling and the vector database directory.
Where a rival wins: teams that already keep their data in Postgres may prefer a Postgres extension, and G2 lists PG Vector among its comparisons. A team that wants an embedded store inside its own process may look at LanceDB instead. Chroma is the rival Weaviate itself courts, with a page for migrating from Chroma Cloud.
Version 1.40 shipped on 7 October 2026 according to the Weaviate release notes read that day, after 1.39 on 4 August and 1.38 on 5 June, and the project supports its latest three minor versions with bug fixes and security patches.
Our call: pick Weaviate for RAG and agent retrieval when an open BSD core you can later self-host matters to you. Pick a Postgres extension or LanceDB if your data already lives in Postgres or must stay in-process, and do not treat the free cluster as a staging copy of production.
Plans, regions, the uptime commitment and the compliance list on this page are read from Weaviate's own pages and documentation. The verdict is Hokai's reading of them, not the provider's claim.