Pinecone

license

proprietary(SaaS)

source

closed

self-host

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

managed

Pinecone-Cloud

Pinecone is available only as a managed service used under a subscription agreement, and none of its source is published. Its BYOC option places the data plane in the customer's own cloud account, but the software stays closed and Pinecone runs the control plane.

• 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.

• https://docs.pinecone.io/guides/production/bring-your-own-cloud recorded 2026-09-28

  The BYOC guide describes a split deployment in which only the data plane runs in the customer's account: "The data plane runs entirely in your cloud account within a dedicated VPC", while "The control plane is managed by Pinecone globally and handles index lifecycle management, authentication, billing, and user management". Searched the body for self-managed, self-host, open source and source code: none.

• https://www.pinecone.io/legal/master-subscription-agreement/ recorded 2026-09-28

  The Master Subscription Agreement grants a right to use the hosted service, not the software: "Pinecone hereby grants to Customer" the "right to use Services in accordance with the Documentation", and the customer may not "attempt to discover or recreate the source code for, Services".

• https://www.pinecone.io/pricing/ recorded 2026-09-28

  The pricing page says "Pinecone runs on fully managed infrastructure that scales with you", and its one deployment alternative is BYOC, which "runs Pinecone in your cloud account and VPC". Searched the body for open source, self-host, on-prem and license: none.

Pinecone publishes no aggregate customer or user count, so the level reads from production-deployment case studies naming specific companies running large vector workloads through its serverless offering and AWS Marketplace. That shows deployment depth at a handful of named accounts rather than breadth across a customer base.

• 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.

• 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.

Pinecone is a distributed, durable vector database run at production scale, and its own documentation describes hosted embedding and reranking models inside the query path: integrated embedding at index time and integrated reranking as part of a search. Hosting the model in the serving path is what sets Vespa and Elasticsearch apart, which puts Pinecone beside them and one above Milvus, whose model utilities run in the client. Its production figures are vendor-reported, not independently benchmarked.

• https://docs.pinecone.io/guides/index-data/create-an-index recorded 2026-10-08

  Create-an-index guide: indexes for "full-text (BM25), semantic (dense vector), sparse-vector, or hybrid search", with "Integrated embedding" index options that "have Pinecone convert it to dense vectors" from source text.

• https://docs.pinecone.io/guides/search/rerank-results recorded 2026-10-08

  Rerank guide: "You can use a hosted model to rerank results as an integrated part of a query"; "To rerank initial results as an integrated part of a query, without any extra steps, use the Search with text operation with the rerank parameter, including the hosted reranking model".

• https://docs.pinecone.io/models/overview recorded 2026-10-08

  Pinecone model gallery: "Browse Pinecone's hosted embedding and reranking model gallery" and "Build end-to-end faster with models hosted by Pinecone."

• https://siliconangle.com/2025/12/01/pinecone-scales-vector-database-support-demanding-workloads/ recorded 2026-10-08

  "5,700 queries per second with a P50 latency of just 26 milliseconds across a database of 1.4 billion 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.

• https://www.infoq.com/news/2025/12/pinecone-drn-vector-workloads/ recorded 2026-10-08

  "Pinecone Introduces Dedicated Read Nodes in Public Preview for Predictable Vector Workloads": dedicated read nodes beside serverless scaling.