Weaviate

Weaviate combines vector similarity search with keyword (BM25) search in a single fused query — 'hybrid search' — rather than requiring you to bolt the two together yourself, and independent July 2026 benchmarking by AIMultiple found its hybrid fusion added a measurable +0.030 nDCG lift over pure vector search on a 50k-document medical/technical corpus. In the same benchmark, Weaviate's single-thread throughput (341 QPS, p95 3.5ms) was mid-pack, but under 32 concurrent worker processes it scaled to the highest peak throughput of the seven engines tested (8,330 QPS) — a workload shape that matters more for production RAG serving than single-thread numbers do. The open-source core is BSD-3-Clause, a genuinely permissive license with no revenue threshold or field-of-use restriction that flips it to source-available, which is not true of every vector database in this space.

Weaviate Cloud publishes real numbers rather than gating everything behind 'contact sales'. The Free tier is $0/month for 100,000 objects, 1GB memory, 10GB disk, and 2,000 embedding requests/day. Flex is pay-as-you-go from $45/month with usage billed separately: vector dimensions from $0.00465 per 1M, storage from $0.12/GiB, backups from $0.0264/GiB. Premium starts at $400/month as a prepaid contract (shared or dedicated deployment) with lower per-unit usage rates (dimensions from $0.002718 per 1M dedicated). Self-hosting the open-source core avoids all of this and is a real option given the BSD license, though it shifts operational cost onto you.

The Python client's v3-to-v4 rewrite changed the import structure, object model, and query API. One team's public account of running Weaviate in production describes spending a day migrating when v4 became the stable default and the v3 compatibility shim began emitting deprecation warnings they couldn't suppress; the same account cites the schema-first design (any metadata structure change requires a schema migration, not just inserting a field) and GraphQL query verbosity as recurring friction, and says Weaviate 'lasted two projects on our stack' before they moved to alternatives for new work. This is one team's experience, not a systematic survey, but it's concrete and specific enough to weigh.

The core repository is active: 16,740 GitHub stars, last push 2026-08-19, and the latest release (v1.38.10) shipped 2026-08-18, two days before this listing was checked. The official Python client was last published to PyPI on 2026-08-13. GitHub's security advisory database lists exactly one advisory against the project — a low-severity denial-of-service issue (CVE-2023-38976, GHSA-8697-479h-5mfp) disclosed and fixed in August 2023 — and OSV's database returns no additional entries. Weaviate publishes a security/trust page referencing SOC 2 and HIPAA compliance and a separate trust portal for documentation, though the compliance claims themselves weren't independently verified here.

Weaviate (the company, formerly SeMI Technologies) is headquartered in Amsterdam and was founded in 2019. It raised a confirmed $50M Series B in April 2023, led by Index Ventures with Battery Ventures, NEA, Cortical Ventures, Zetta Venture Partners and ING Ventures participating (via PRNewswire). Multiple third-party funding trackers (Tracxn, SalesTools) report a further $50M Series C in October 2025 at a $200M valuation led by Battery Ventures and Zetta Venture Partners, but no primary press release or wire report for that round could be located — treat the Series C figure as reported-but-unconfirmed rather than established.

Weaviate holds a 4.6/5 rating on G2, but from a small sample — 29 reviews as of this check. That's too thin to treat as a representative signal either way; it's evidence of no major dissatisfaction among reviewers, not proof of broad enterprise satisfaction.