At a glance
| Product type | Standalone open source vector database server, plus Weaviate Cloud managed service |
| Vector index types | HNSW (default), flat, dynamic (flat to HNSW), HFresh (cluster-based, disk-resident) |
| Hybrid search | Yes: BM25 plus vector, relativeScoreFusion or rankedFusion, alpha weighting |
| Metadata filtering | Yes: filters on object properties |
| Embedding integration | Built-in model provider modules (API-based and local) and Weaviate Embeddings |
| Scaling and multitenancy | Sharding and replication; native multi-tenancy with a shard per tenant |
| Deployment (managed or self-hosted) | Self-hosted (Docker, Kubernetes) or Weaviate Cloud (Free, Flex, Premium shared or dedicated) |
| Licensing and pricing model | BSD-3-Clause core; cloud billed per vector dimension, storage and backup |
Weaviate fits teams that want an open source vector database with hybrid search and multi-tenancy built in, plus the option of a managed service. The forever-free cloud plan is small (100,000 objects), so production on Weaviate Cloud means Flex from $45 a month or a Premium contract, billed largely on stored vector dimensions.
Overview
Weaviate is an AI-native vector database developed by the company Weaviate and published on GitHub, mostly under the BSD 3-Clause licence. It is a standalone database server written in Go that you run with Docker or on Kubernetes, or use as a managed service through Weaviate Cloud. Data is organised in collections of objects that carry properties and one or more named vectors, queried through GraphQL, REST and gRPC APIs and client libraries.
The default vector index is HNSW, with a flat index for small collections and multi-tenancy, a dynamic index that switches from flat to HNSW past a threshold (experimental, v1.25 and later), and HFresh (v1.36 and later), a cluster-based index that keeps most data on disk. Weaviate can call embedding, generative and reranker models from many providers itself, so applications can insert and query plain text. Multi-tenancy is native: each tenant lives in its own shard and can be active, inactive or offloaded to cloud storage.
Key features
- Vector index choice per collection: HNSW (default), flat, dynamic and the disk-oriented HFresh index
- Hybrid search that runs BM25 and vector search in parallel and fuses scores (relativeScoreFusion by default), weighted by an alpha parameter
- Native multi-tenancy with one shard per tenant and ACTIVE, INACTIVE and OFFLOADED tenant states
- Model provider integrations for embeddings, generative models and rerankers, both API-based (OpenAI, Cohere, AWS, Google, Voyage AI and others) and local (Ollama, Hugging Face Transformers)
- Weaviate Embeddings, a hosted embedding service billed per million tokens
- Multi-vector support (ColBERT-style) on HNSW and HFresh, with optional MUVERA encoding
- Quantization to cut memory use, including rotational quantization (RQ)
Integrations and ecosystem
Pros and cons
Pros
- Permissive BSD-3-Clause licence for the core database, so it can be self-hosted at no licence cost
- Hybrid BM25 and vector search with documented fusion methods
- Native, shard-per-tenant multi-tenancy with offloading of inactive tenants
- Wide choice of embedding, generative and reranker provider integrations
Cons
- The default HNSW index is built in memory, so large self-hosted deployments need RAM planning or compression
- The cloud Free plan is limited to 100,000 objects and 1 collection
- Cloud billing by vector dimensions varies with index type, compression and region, so estimates need the calculator
- Part of the repository (the wl/ folder) is under a commercial licence, so check which components you use
Pricing
Self-hosting the open source database costs nothing in licence fees. Weaviate Cloud has three plans. Free is $0 with no stated time limit: 1 cluster, 100,000 objects, 1 GB memory, 10 GB disk, 1 collection and up to 3 tenants. Flex is pay-as-you-go from $45 a month, with no commitment. Premium (shared or dedicated) is a prepaid contract from $400 a month, priced by sales. Usage is billed per 1M vector dimensions (from $0.00465 on Flex, $0.003875 Premium shared, $0.002718 Premium dedicated), storage per GiB (from $0.12, $0.10 and $0.33) and backup per GiB; dimension rates vary by index type, compression and region, and the price book changed for usage from 1 September 2026. Weaviate Embeddings is billed per 1M tokens (from $0.025). Prices use a $ sign without a currency code.
Pricing checked on the vendor's official page on 7 November 2026. Prices change; confirm before buying.
Who it is best for
Teams building RAG or search applications who want an open source vector database with hybrid search and multi-tenancy, and the option to move between self-hosted and managed.
Typical use cases
- Hybrid keyword and semantic search over documents or products
- Multi-tenant SaaS applications with one isolated shard per customer
- Retrieval-augmented generation using built-in embedding and generative modules
- Multimodal and multi-vector retrieval such as ColBERT or ColPali
Who should look elsewhere
- Teams that want no infrastructure choices at all and prefer a closed managed service such as Pinecone.
- Small single-process prototypes that are better served by an embedded library such as Chroma or LanceDB.
Verdict
Weaviate is a well-rounded open source vector database whose strengths are hybrid search, native multi-tenancy and built-in model integrations. Self-hosting is free but needs memory planning for HNSW, while Weaviate Cloud pricing is usage-based on vector dimensions, so test both cost models against your data size before choosing.
Frequently asked questions
Is Weaviate open source?
Yes. Most of the weaviate/weaviate repository is under the BSD 3-Clause licence, though files in its wl/ folder are under a commercial licence. Weaviate Cloud is a separate paid managed service.
Does Weaviate Cloud have a free tier?
Yes. The Free plan has no stated time limit and allows 1 cluster, 100,000 objects, 1 GB memory, 10 GB disk, 1 collection and up to 3 tenants. Paid Flex starts from $45 a month.
Does Weaviate support hybrid search?
Yes. It runs BM25 keyword search and vector search together and fuses the scores, with alpha controlling the balance (0 is keyword only, 1 is vector only; the server default is 0.75).
How does Weaviate handle multi-tenancy?
You enable it per collection, and each tenant is stored in its own shard. Tenants can be ACTIVE, INACTIVE (kept on disk) or OFFLOADED to cloud storage through the S3 offload module.
Sources
- Weaviate Cloud pricing
- Weaviate vector index concepts
- Weaviate hybrid search concepts
- Weaviate multi-tenancy documentation
- Weaviate model provider integrations
- weaviate/weaviate GitHub repository (licence)
Checked November 2026.
Listed in: Vector Databases & AI Search.
How we research reviews and what star ratings mean: our editorial method.