Vital signs
Why it's on the table
On the table, Vector Store (Vd) is seat 14 of 58, in the Knowledge & Memory family. It is a stable element — the category has settled, choosing is cheap, and switching is rare. Pick a holder and move on; this is not where your decision budget should go. It is optional: plenty of companies run without it — until a specific trigger (scale, regulation, cost, or customers) makes it essential for them. It sits in the lowest paid band — lunch money against the hours it returns.
Vector Store: the top 5 — v2026.Q3
Edition v2026.Q3 · ranking, pricing and status verified 2026-08-06.
1pgvectoropen source (Andrew Kane + community)
Free, PostgreSQL-licensed · runs in any Postgres — RDS, Supabase (free tier), Neon, Cloud SQLBest for Every startup that already runs Postgres and holds under ~10–50M vectors — which is roughly 95% of them.
One extension turns the database you already operate into your vector store — same backups, same transactions, joins against your real data, zero new vendor. v0.8.2 (Feb 25, 2026) keeps shipping: iterative index scans, half-precision and sparse vectors, binary quantization. Documented migrations from Pinecone to pgvector at 1/10th the monthly cost are the category's defining production story.
Watch Vanilla HNSW strains past ~50M vectors — you'll want pgvectorscale (DiskANN) or a graduation plan. Index build memory and write-heavy churn need real Postgres operational skill; 'free' is not 'zero-ops'.
v0.8.2 released Feb 25, 2026; 22.4k GitHub stars (Aug 2026) [src] · Production migration reports: 500K daily queries / 8M vectors at ~$200/mo on pgvector vs $2,000+/mo on Pinecone (2025–26 roundup) [src] · With pgvectorscale: 471 QPS at 28ms p95 on 50M embeddings in benchmarks (vendor benchmark, Timescale/Tiger Data) [src]2Turbopufferturbopuffer, Inc.
Usage-based · Launch $16/mo min · Scale $256/mo min (HIPAA BAA, SSO) · Enterprise ≥$4,096/mo (single-tenant, BYOC)Best for Multi-tenant AI products with millions of per-user namespaces — the exact shape of agent memory and per-customer document search.
Search built natively on object storage with an NVMe/memory cache: ~10x cheaper economics, sub-10ms p50 hot queries, 1T+ documents and 250M+ namespaces in production. The customer list is the 2026 AI stack itself — Cursor (its largest, 1T+ code chunks), Notion (10B+ vectors, migrated off Pinecone Oct 2024), Anthropic, Linear, Harvey. Sacra puts it at $100M annualized revenue by March 2026 on under $1M of primary capital raised — the most efficient company in this table, possibly in any table.
Watch Cold namespaces pay object-storage latency (hundreds of ms) on first hit. Closed source, no self-host below Enterprise BYOC, and a small team carrying very large customers. Revenue figure is single-source (Sacra).
$100M annualized revenue (Mar 2026), up from $75M at end-2025; <$1M primary capital raised; seed with Thrive Capital Dec 2025 [src] · 1T+ documents, 250M+ namespaces, 10M+ writes/sec globally; customers incl. Cursor, Notion, Anthropic, Linear (Aug 2026) [src] · Entry at $16/mo minimum usage; Enterprise from $4,096/mo (Aug 2026) [src]3PineconePinecone Systems
Starter free (2GB) · Builder $20/mo flat · Standard $50/mo min + usage · Enterprise $500/mo minBest for Teams that want a zero-ops, pure-play managed vector database with mature enterprise controls and don't want to think about infrastructure at all.
The category's brand name, still: 5,000+ customers, serverless architecture, integrated inference/embedding endpoints, SOC 2/HIPAA, BYOC. The 2025 reset — founder Edo Liberty stepping to Chief Scientist, three-time founder Ash Ashutosh taking CEO in September 2025, and a new $20/mo Builder tier — is a deliberate push down-market and toward 'knowledgeable AI' platform positioning.
Watch The cautionary tale of the category: flagship logos (Notion) left over cost, mid-2025 press cycles covered sale-exploration chatter (the new CEO says acquisition is 'definitely not on the table'), and it has raised nothing since 2023's $100M at $750M. At startup scale you are paying a premium for convenience pgvector gives you free.
4QdrantQdrant (Berlin)
OSS free (Apache-2.0, Rust) · Cloud free 1GB cluster · Standard usage-based · Premium/Private Cloud customBest for Self-hosters and regulated teams who need heavy metadata filtering, hybrid search, and full control from laptop to 1B+ vectors.
The strongest open-source pure-play in 2026: Rust core, best-in-class filtered search, quantization that cuts memory 4–32x, and hybrid/multi-vector support. A $50M Series B (Mar 12, 2026, Advance Venture Partners; $87.8M total) explicitly aimed at 'agentic retrieval' keeps it independently funded while rivals consolidate. TripAdvisor runs 1B+ vectors on it.
Watch Cloud pricing beyond the free tier is quote-driven — less transparent than Turbopuffer or Chroma. Feature surface (embeddings, rerankers, agents) is expanding fast; watch for focus dilution.
5ChromaChroma (San Francisco)
OSS free (Apache-2.0) · Cloud usage-based: $0.33/GiB-mo storage, $2.50/GiB writes · Starter $0 + $5 credit · Team $250/moBest for AI engineers who want retrieval running in five minutes — local in a notebook today, serverless Chroma Cloud with the same API tomorrow.
The developer-experience pick and the default vector store scaffolded by AI coding agents: 28k GitHub stars, single-node simplicity, vector + full-text + hybrid in one API. Chroma Cloud (2025, Rust-rewritten, object-storage-backed) added a genuinely cheap transparent usage model, and the team's retrieval research (context-rot, generative benchmarking) earns real credibility.
Watch Thinnest capitalization of the five ($18M seed, Apr 2023 — no announced round since) and the least enterprise track record at billion scale. Single-node OSS deployments hit walls that force the Cloud migration decision.
Vector Store: the top 8 compared
Edition v2026.Q3 · ranking, pricing and status verified 2026-08-06.
| Tool | Entry price | Open source | Deploy | Hybrid (BM25+vector) | Sweet-spot scale | Architecture | Enterprise |
|---|---|---|---|---|---|---|---|
| pgvector | Free | Yes (PostgreSQL lic.) | In your Postgres, anywhere | Via Postgres FTS | <10–50M vectors | HNSW/IVFFlat in-database | Inherits your Postgres stack |
| Turbopuffer | $16/mo min | No | Managed · BYOC at Enterprise | Yes (native BM25) | 10M–100B+, many namespaces | Object storage + NVMe cache | SOC 2 · HIPAA · CMEK · 99.95% SLA |
| Pinecone | Free · $20/mo Builder | No | Managed · BYOC at Enterprise | Yes (sparse+dense) | 1M–1B+ | Serverless, separated storage/compute | SOC 2 · HIPAA · RBAC · audit logs |
| Qdrant | Free OSS · free 1GB cloud | Yes (Apache-2.0) | Self-host · cloud · hybrid · private | Yes | 10M–1B+ | Rust, HNSW + quantization | SSO · VPC · custom SLAs |
| Chroma | Free OSS · $0 + usage cloud | Yes (Apache-2.0) | Local · self-host · serverless cloud | Yes | <100M | Rust core on object storage | SOC 2 · single-tenant on Team+ |
| Weaviate | Free sandbox · Flex $45/mo | Yes (BSD-3) | Self-host · serverless · dedicated | Yes | 1M–1B | Go, HNSW, modules ecosystem | HIPAA (AWS) · PrivateLink · CMEK |
| Milvus / Zilliz | Free OSS · serverless $0 · dedicated ~$126/GB-mo | Yes (Apache-2.0) | Self-host · cloud · BYOC | Yes (v2.5+) | 100M–10B+ | Distributed, tiered storage, GPU option | 99.95% SLA · HIPAA-eligible · BC tier |
| S3 Vectors | Pure usage, no minimum | No | AWS only | No (ANN only) | Cold/batch, up to 2B per index | Native S3 vector buckets | Standard AWS compliance stack |
How to choose your vector store
- If you already run Postgres and hold under ~10M vectors (you almost certainly do)
- pgvector. One extension, zero new vendors, joins against your real data. Revisit only when p95 latency or index-build pain says so.
- If you're building a multi-tenant AI product — per-user memory, per-customer document search — or your dedicated-DB bill crossed four figures
- Turbopuffer. Namespaces-on-object-storage is purpose-built for that shape, and it's the migration Notion and Cursor already validated.
- If procurement wants a managed pure-play with every compliance box ticked, and nobody wants to run infrastructure
- Pinecone Standard ($50/mo min) — accepting that you're paying the convenience premium the rest of this page is about avoiding.
- If you need self-hosting, data residency, or heavy metadata-filtered hybrid search at 100M+ vectors
- Qdrant OSS, with their cloud or hybrid cloud when ops outgrows you. Milvus/Zilliz if you're genuinely at billions.
- If your vectors are cold, batch-queried, and already in AWS — archives, logs, agent episodic memory
- S3 Vectors: no minimums, up to ~90% cheaper, sub-second cold queries. Pair it with a hot tier; it is not a low-latency serving store.
Vector Store: the whole field
19 more tools tracked in this category, including 6 dead, renamed, or sunsetting — a reference that hides the graveyard isn't one. Verified 2026-08-06.
| Tool | Maker | What it is | Entry | Status |
|---|---|---|---|---|
| Weaviate | Weaviate B.V. | The near-miss for the top 5 — mature OSS with module ecosystem and Query Agent; $68M raised, Ricoh strategic investment Mar 13, 2026; momentum flatter than Qdrant's | free sandbox · Flex $45/mo | active |
| Milvus / Zilliz Cloud | Zilliz | The billion-scale OSS heavyweight: 40k GitHub stars (Dec 2025), 10k+ enterprise deployments (NVIDIA, Salesforce, eBay); tiered storage cut costs 50% in v2.6; Milvus Lake lakehouse coming | free · dedicated from ~$16/M vectors/mo | active |
| LanceDB | LanceDB (Lance format) | $30M Series A (Jun 2025) to build the 'multimodal lakehouse' — embedded OSS + columnar Lance format; Netflix, ByteDance Volcano Engine, CrewAI (2B+ agent executions) | free OSS · cloud usage-based | active |
| pgvectorscale | Tiger Data (ex-Timescale) | DiskANN + statistical binary quantization on top of pgvector — the extension that stretches Postgres past 50M vectors; 0.9.0 added Postgres 18 support | free (PostgreSQL lic.) | active |
| Amazon S3 Vectors | AWS | GA Dec 2, 2025 — vector buckets native to S3, 2B vectors/index, ~100ms warm queries, 'up to 90% cost reduction'; the price floor under the whole category | pure usage, no minimum | active |
| MongoDB Atlas Vector Search | MongoDB | Vector search absorbed into the document DB — plus the Voyage AI embedding/reranker acquisition (Feb 2025) for an integrated retrieval stack | free tier · Atlas usage | active |
| Elasticsearch / OpenSearch | Elastic / OpenSearch Software Foundation | Incumbent search engines with dense-vector + BM25 hybrid; the default when you already run ELK — and Elastic leadership has been openly dismissive of standalone vector DBs as a business | free self-host · cloud from ~$95/mo | active |
| Redis (Vector Sets / Query Engine) | Redis Ltd. | In-memory vector search inside Redis 8 — lowest latency for small hot sets and semantic caching; RAM economics cap the scale | free · Redis Cloud from $5/mo | active |
| Vespa | Vespa.ai (Yahoo spin-out) | Big-league serving engine (search + recsys + vectors) used by Spotify and Yahoo properties; overkill below hundreds of millions of documents | free OSS · cloud usage | active |
| Supabase (pgvector hosting) | Supabase | The biggest on-ramp to pgvector — free tier includes it; the reason 'just use Postgres' is a two-minute decision for new projects (platform itself → element Ba) | free tier · Pro $25/mo | active |
| Neon (pgvector serverless) | Neon → Databricks | Serverless Postgres with pgvector; acquired by Databricks May 2025 (~$1B reported) — vectors ride into the lakehouse ecosystem | free tier | acquired |
| Astra DB / DataStax | DataStax → IBM | Cassandra-based vector DB; DataStax acquisition by IBM closed 2025 — vector search folded into watsonx orbit | free tier · usage | acquired |
| Vectara | Vectara | RAG-as-a-service platform rather than a raw vector store — retrieval, reranking, and grounded generation in one API; adjacent but often cross-shopped | free tier · usage | active |
| Marqo | Marqo | End-to-end embedding search (brings its own inference); niche but differentiated in e-commerce search | free OSS · cloud usage | active |
| Faiss | Meta | The foundational ANN library — not a database; still what many in-house vector systems are built on | free (MIT) | active |
| Deep Lake | Activeloop | Tensor/vector data lake for multimodal ML; mindshare faded as LanceDB took the format-centric lane | free OSS · cloud tiers | fading |
| MyScale | MyScale (ClickHouse-based) | SQL vector DB on ClickHouse; quiet through 2025–26, community momentum stalled | free tier | fading |
| Annoy | Spotify | The 2015-era ANN library; superseded by Spotify's own Voyager and modern HNSW implementations — historical interest only | free | dead |
| pg_embedding | Neon | Neon's rival Postgres extension, deprecated in 2023 in favor of pgvector — early evidence the ecosystem consolidates on one Postgres answer | — | dead |
Vector Store: the category in numbers
Edition v2026.Q3 · ranking, pricing and status verified 2026-08-06.
- Vector DB market ~$2.55B (2025) → $15.1B by 2034, 22.3% CAGR; NLP workloads the largest segment [src]
- Turbopuffer: $100M annualized revenue (Mar 2026) on <$1M primary capital — Cursor its largest customer at 1T+ indexed code chunks [src]
- Qdrant $50M Series B (Mar 12, 2026, AVP; $87.8M total) — the only large pure-play round of 2026, aimed at agentic retrieval [src]
- AWS S3 Vectors GA Dec 2, 2025: 2B vectors/index (40x preview), ~100ms warm queries, 'up to 90%' cost reduction — commoditizing cold vector storage [src]
- Consolidation into incumbents: SQL Server 2025, MongoDB Atlas, Redis 8, Cassandra, Oracle all ship vector search; Pinecone swapped CEOs (Sep 2025) after sale-exploration press; Neon and DataStax absorbed by Databricks and IBM (2025) [src]
- Milvus passed 40k GitHub stars with 10k+ enterprise deployments (Dec 18, 2025); Gartner projects vector capabilities in 30% of enterprise apps by 2026 [src]
Vector Store: method & sources
Ranking criteria: total cost of ownership at startup scale (<50M vectors), production adoption by AI-native companies as of Aug 2026, operational simplicity, and survivability of the vendor. Conflicts resolved: pgvector's GitHub README rendered a '0.8.5' string but the canonical CHANGELOG's latest entry is 0.8.2 (Feb 25, 2026) — we use the changelog. pgvectorscale 0.9.0's GitHub tag renders an ambiguous date (page said Nov 4, 2024, impossible for a Postgres 18 feature; actual is almost certainly Nov 2025) — we avoid asserting the date. Turbopuffer's $100M revenue is Sacra-only (no company confirmation): kept, flagged. The 'Pinecone exploring sale' claim traces to mid-2025 press chatter and aggregator retellings; the on-record fact is the new CEO's denial (TechTarget, Sep 2025). Elastic-CEO 'vector DBs were never a business' quote circulates via aggregators without a clean primary link — treated as sentiment, not sourced fact. Market-size figures vary by firm ($2.5–3B for 2025/26); we cite GM Insights as a representative mid-range. Adjacent elements: agent memory APIs (Mem0, Zep, Letta) → Mm · Memory Layer; RAG-as-a-service and research tools → Re; wikis your agents read → Kw. Embedding models (OpenAI, Voyage, Cohere) are a model-layer choice, not a store, and are deliberately excluded. Ranking criteria: verified commercial traction, independent satisfaction surveys, agent benchmarks, and founder-fit (price floor, lock-in, surfaces). Editorial, never paid — the charter. Machine-readable twin: vd.json.
All sources (21)
- https://github.com/pgvector/pgvector/blob/master/CHANGELOG.md
- https://www.postgresql.org/about/news/pgvector-082-released-3245
- https://turbopuffer.com/
- https://turbopuffer.com/pricing
- https://sacra.com/c/turbopuffer/
- https://www.pinecone.io/pricing/
- https://www.prnewswire.com/news-releases/pinecone-founder-edo-liberty-to-spearhead-pinecones-growing-ai-ambitions-appoints-ash-ashutosh-as-ceo-to-expand-vector-database-market-leadership-302549334.html
- https://www.techtarget.com/searchdatamanagement/news/366631366/Vector-database-vendor-Pinecone-eyes-future-under-new-CEO
- https://www.techtarget.com/searchdatamanagement/news/366640132/Qdrant-raises-50M-in-funding-to-fuel-vector-database-growth
- https://qdrant.tech/pricing/
- https://www.trychroma.com/pricing
- https://github.com/chroma-core/chroma
- https://weaviate.io/pricing
- https://zilliz.com/pricing
- https://finance.yahoo.com/news/milvus-surpasses-40-000-github-010000562.html
- https://www.lancedb.com/blog/series-a-funding
- https://aws.amazon.com/about-aws/whats-new/2025/12/amazon-s3-vectors-generally-available/
- https://github.com/timescale/pgvectorscale/releases
- https://www.gminsights.com/industry-analysis/vector-database-market
- https://medium.com/data-science-collective/vector-databases-are-dying-heres-the-production-evidence-8c17b54687e2
- https://www.ricoh.com/release/2026/0616_1
Our take
Start with pgvector inside the database you already run. Graduate only when scale forces you to.
Combines with
This is element 14 of 58. The table is versioned quarterly — when a tool loses its seat, the changelog records the succession.
Explore the full table →