Pinecone

license

proprietary(SaaS)

source

closed

self-host

none(BYOC runs Pinecone-managed in customer VPC, not OSS)

managed

Pinecone-Cloud

Fully proprietary managed vector DB; no open source core (contrast Milvus/Qdrant which are OSI-licensed). BYOC is still closed-source software run in customer cloud.

• https://www.pinecone.io/pricing/ recorded 2026-08-13

  Tiers Starter, Builder, Standard and Enterprise, plus BYOC: "Bring-your-own-cloud (BYOC) runs Pinecone in your cloud account and VPC. Pinecone does not need SSH, VPN, or inbound network access to operate the system." Enterprise adds a 99.95% uptime SLA, private endpoints and customer-managed keys. The phrase "open source" does not appear on the page.

• https://docs.pinecone.io/guides/get-started/overview recorded 2026-08-13

  Pinecone documented as a fully managed vector database reached through an API key; no source release or self-hostable build offered.

The leading managed vector database. Production deployments are cited at 1.4 billion vectors and 5,700 queries per second for a single e-commerce customer, and 135 million vectors at 2,200 queries per second load-tested for another, with a broad enterprise and developer base reached through serverless and AWS Marketplace pay-as-you-go. No single hard user count is published, so the level is a judgment across aggregate production usage. Some churn is visible too - Notion moved off it over cost.

• https://siliconangle.com/2025/12/01/pinecone-scales-vector-database-support-demanding-workloads/ recorded 2026-08-13

  Two named-shape customer deployments: 600 QPS at 45ms over 135 million vectors, load-tested to "2,200 queries per second with a P50 latency of just 60 milliseconds"; and "5,700 queries per second with a P50 latency of just 26 milliseconds across a database of 1.4 billion vectors". No user or customer count.

• https://aws.amazon.com/marketplace/pp/prodview-xhgyscinlz4jk recorded 2026-08-13

  Pinecone listed on AWS Marketplace with pay-as-you-go availability. Distribution breadth only; no usage figure.

Level with Milvus, which is the 4 rung on the storage feature matrix, which bands on how much of the persist-index-retrieve path a product provides itself, at what scale and over how many retrieval modes: a complete durable system operated at scale, with integrated sparse-dense hybrid search, serverless auto-indexing and dedicated read nodes. It does not reach the 5 rung, which is for a platform that hosts and evaluates ranking or embedding models inside the serving path, because nothing in the recorded evidence shows that. Being a closed system its figures are vendor-reported and no independent submission exists: ann-benchmarks.com lists 38 algorithms and Pinecone is not among them, so the absence is a confirmed finding rather than an assumption, which is what holds confidence at medium. Rebanded on this matrix when the product moved out of agent_tools_protocols; the band did not move. Capability is independent of openness here - this product scores 1/closed and still reaches the same rung as two open engines.

• https://www.infoq.com/news/2025/12/pinecone-drn-vector-workloads/ recorded 2026-08-18

  Dedicated read nodes for predictable vector workloads; serverless scaling.

• https://siliconangle.com/2025/12/01/pinecone-scales-vector-database-support-demanding-workloads/ recorded 2026-08-18

  "5,700 queries per second with a P50 latency of just 26 milliseconds across a database of 1.4 billion vectors"; and 600 QPS at 45ms over 135 million vectors, load-tested to "2,200 queries per second with a P50 latency of just 60 milliseconds". Vendor-sourced figures reported by a trade outlet, not an independent harness.