Skip to content

Weaviate

Go vector database with first-class hybrid keyword-and-vector search, built-in embedding modules and multi-tenancy, self-hosted or cloud.

Reviewed by Sheela · Updated

View on GitHub Open source

Our rating

4.2/5

Hybrid keyword-and-vector search is still Weaviate's best argument, but its newest operational features now sit behind a licence key.

What it is

Weaviate is a vector database written in Go. It stores objects with their embeddings and can create those embeddings itself through modules that call model providers, so simple ingestion needs no separate pipeline. Hybrid search runs BM25 keyword scoring and vector similarity in one query and blends the results.

It supports multi-tenancy, several compression methods (including 4-bit rotational quantisation), replication and namespaces, which became generally available in 1.40.0 on 7 October 2026. Patch releases continue on four lines: 1.40.2, 1.39.13 and 1.38.22 on 22 October and 1.36.24 on 23 October.

The licence changed in September 2026. Code outside a new wl directory remains BSD-3-Clause, but wl holds enterprise features, including namespaces and deduplicated backups, under a separate Weaviate licence that needs a key. Weaviate Cloud is the managed service.

Our review

Weaviate's case rests on hybrid search. One query blends BM25 keyword scores with vector similarity under a weight you can tune, so exact strings such as product codes and names are not blurred away by embeddings. Reviewers on Product Hunt name this as the reason they chose it.

The licence now needs reading. Since September 2026 the repository is open core, and GitHub no longer recognises it as a single standard licence. Namespaces, which add control-plane and data isolation between users on a shared cluster, reached general availability in 1.40 under a change titled "Make namespaces a feature requiring license". The LICENSE file adds that the BSD terms grant no right to use those features or circumvent the key, and it publishes no evaluation terms.

Self-hosters should also read the telemetry page. Weaviate collects telemetry by default: a daily report with server version, modules, object and collection counts and, from 1.33, the AWS account, GCP project or Azure subscription ID where the cloud metadata endpoint is reachable. The docs say this may be used to identify potential commercial customers and shared with third parties for that purpose. DISABLE_TELEMETRY=true turns it off.

Weaviate Cloud bills by stored vector dimensions. At the Flex list rate of $0.00465 per million, 10 million 1,536-dimension vectors come to about $71 a month before storage at $0.12 per GiB, and SSO arrives only on Premium, from $400 a month. Rates vary by index type and compression, with separate pricing for 4-bit rotational quantisation since 1 September, so compression is a billing decision as much as a technical one.

Where it earns its place

  • A search engineer at an online retailer with two million SKUs

    Return results that match both exact part numbers and loosely worded queries

    Hybrid search blends BM25 and vector scores in one query, so a model number and a vague description both find the right product. The BSD core is free to self-host; budget for the infrastructure and for an engineer who owns schema design.

  • A SaaS platform team serving hundreds of customer workspaces

    Keep each customer's data isolated on shared infrastructure

    Multi-tenancy is in the open core; namespaces add control-plane isolation but need a licence key, and the price is not published. Settle that cost before designing around namespaces.

  • A startup CTO choosing between self-hosting and Weaviate Cloud

    Forecast the monthly bill for a 10-million-vector RAG index

    On Flex, 10M vectors at 1,536 dimensions is about 15.4 billion dimensions, roughly $71 a month at $0.00465 per million before compression, plus $0.12 per GiB stored and a $45 minimum. Customers who require SSO push the bill to Premium, from $400 a month.

Before you commit

Setup
Docker Compose or Kubernetes for self-hosting, or a free Weaviate Cloud sandbox capped at 100,000 objects and one collection. The Python client, weaviate-client 4.24.0, needs Python 3.10+. The schema-first model means planning collections and vectoriser modules before loading data.
Integrations
Vectoriser modules that call model providers, Weaviate Embeddings at $0.025 to $0.065 per million tokens, a Query Agent at $30 per organisation per month with 4,000 requests, a REST search API and official client libraries.
Data and privacy
Self-hosted data stays on your servers, but telemetry is on by default and includes cloud account IDs; set DISABLE_TELEMETRY=true. Weaviate Cloud is SOC 2 Type II audited; HIPAA covers only Dedicated on AWS, and SSO/SAML starts at Premium. GitHub lists one advisory, a low-severity denial of service from 2023.
Cost at scale
Flex: $0.00465 per million dimensions, $0.12 per GiB stored, $45 minimum, 99.5% SLA. Premium Shared: $0.003875 per million dimensions and $0.10 per GiB, 99.9% SLA, from $400 a month. Premium Dedicated: $0.002718 per million dimensions, 99.95% SLA. Enterprise licence keys for self-hosting are priced on request.
Limits
805 open issues and pull requests. The free cloud tier allows one collection and three tenants with no backups; Flex's severity-1 response is one business day. Enterprise features in wl cannot be used without a key.

What's good

  • Hybrid BM25 and vector search in one query
  • Built-in embedding modules, so ingestion needs no separate pipeline
  • Several compression options, including 4-bit rotational quantisation
  • Patch releases across four supported minor versions

What's not

  • Namespaces and deduplicated backups need a commercial licence key
  • Telemetry on by default, including cloud account IDs
  • SSO only on Premium cloud; HIPAA only on Dedicated
  • Dimension-based cloud billing is hard to forecast

Best for

Teams building search or RAG where exact keywords and semantic similarity both matter.

Alternatives

Weaviate: common questions

Is Weaviate worth it?
Hybrid keyword-and-vector search is still Weaviate's best argument, but its newest operational features now sit behind a licence key. We rate it 4.2 out of 5.
How much does Weaviate cost?
Core is BSD-3-Clause; some enterprise features need a paid licence key. Weaviate Cloud: Free (100,000 objects), Flex pay-as-you-go with a $45/month minimum, Premium from $400/month on a prepaid contract.
What are the downsides of Weaviate?
Namespaces and deduplicated backups need a commercial licence key; telemetry on by default, including cloud account IDs; sSO only on Premium cloud; HIPAA only on Dedicated; dimension-based cloud billing is hard to forecast.
Who is Weaviate best for?
Teams building search or RAG where exact keywords and semantic similarity both matter.
How much does Weaviate cost for a team or in production?
Flex: $0.00465 per million dimensions, $0.12 per GiB stored, $45 minimum, 99.5% SLA. Premium Shared: $0.003875 per million dimensions and $0.10 per GiB, 99.9% SLA, from $400 a month. Premium Dedicated: $0.002718 per million dimensions, 99.95% SLA. Enterprise licence keys for self-hosting are priced on request.
Is Weaviate safe for company data?
Self-hosted data stays on your servers, but telemetry is on by default and includes cloud account IDs; set DISABLE_TELEMETRY=true. Weaviate Cloud is SOC 2 Type II audited; HIPAA covers only Dedicated on AWS, and SSO/SAML starts at Premium. GitHub lists one advisory, a low-severity denial of service from 2023.
What are the best Weaviate alternatives?
The closest alternatives we have reviewed are Qdrant, Milvus, pgvector and Chroma.

Answers checked Oct 23, 2026.

Make Weaviate? Show this rating on your siteWeaviate rated 4.2 out of 5 on ReviewRamble