What it is
Pinecone is a hosted vector database: you send it embeddings (number lists that represent the meaning of text or images) and it returns the closest matches for a query. There is nothing to install or run. In 2026 it became fully serverless; the older pod-based indexes are now legacy and cannot be created from the latest SDKs.
Around the database it sells hosted embedding and reranking models, Pinecone Assistant (a managed RAG service) and Nexus, a knowledge engine for agents that reached general availability on 6 August 2026. Full-text search became generally available on 2 September 2026, and Dedicated Read Nodes (reserved capacity for read-heavy indexes) in April.
Enterprise customers can run Pinecone inside their own AWS, Azure or Google Cloud account (bring your own cloud), generally available since 25 August 2026.
Our review
Pinecone sells the absence of a database team. There is no server, index tuning or upgrade path to manage, and the 2026 product closes most of the gaps larger buyers complained about: full-text search went generally available on 2 September, Dedicated Read Nodes in April and bring-your-own-cloud on AWS, Azure and Google Cloud on 25 August. The price of that convenience is a closed, managed-only service. Pinecone retired its Docker emulator, Pinecone Local, on 7 October 2026, so there is not even a local stand-in any more, and leaving means re-indexing elsewhere.
The bill is shaped by reads. Standard charges $16 to $18 per million read units and Enterprise $24 to $27, on top of $0.33 per GB a month for storage, and since 1 September egress beyond 100 GB is metered at $0.10 per GB. That is cheap for a read-light RAG app and expensive for a busy one. Encore, which sells Postgres hosting and so has a stake, puts 1 to 5 million vectors at $100 to $500 a month on Pinecone against little extra for pgvector on an existing database.
Tier boundaries matter more than the headline price. SSO and role-based access arrive on Standard; HIPAA is a $190-a-month add-on there and included only on Enterprise, which is also the only tier with private endpoints, customer-managed keys, audit logs, SCIM and a 99.95% uptime SLA. The $20 Builder plan blocks reads once it hits its quotas rather than billing overages, a real cap but also a way for a production app to stop answering.
Performance figures are vendor claims. Pinecone's product page quotes p50 latency of 16 ms and p99 of 33 ms for a dense index of 10 million records in one namespace, without saying what hardware or query load produced them. The more telling data point is commercial: The Information reported in 2025 that Pinecone explored a sale after losing Notion as a flagship customer, and Notion now appears on turbopuffer's customer list. Founder Edo Liberty handed the chief executive role to Ash Ashutosh in September 2025, and no sale has been announced since.
Where it earns its place
A three-engineer startup shipping a RAG feature on AWS
Add semantic search without hiring for database operations
Start on the $20 Builder plan with 10 GB storage and 2 million read units a month; reads stop at the cap rather than running up a bill. Moving to Standard costs at least $50 a month, then $16 to $18 per million reads.
A head of engineering at a US health-tech company
Store embeddings of patient notes under a BAA
Standard plus the $190-a-month HIPAA add-on covers the BAA, but private endpoints, customer-managed keys and audit logs need Enterprise at a $500 monthly minimum. Budget for Enterprise if security review asks for any of the three.
A platform team serving 20,000 customer workspaces
Give each customer its own isolated index
Standard allows 100,000 namespaces per index and Enterprise 1,000,000, with data-plane requests limited to 100 per second per namespace. Namespace aliases, added in September 2026, let the team repoint a tenant to a rebuilt namespace without a redeploy.
Before you commit
- Setup
- Create an API key, install the Python, Node or Go SDK and create an index; the latest SDKs target API version 2026-07 and no longer create pod-based indexes.
- Integrations
- Python, Node.js, Go and other SDKs, a REST API, a Terraform provider, MCP servers, hosted embedding models and Cohere and BGE rerankers.
- Data and privacy
- SOC 2 Type II, ISO 27001 and GDPR on paid tiers; HIPAA as a $190-a-month add-on on Standard or included on Enterprise; data encrypted with AES-256 at rest; BYOC runs inside your own cloud account on Enterprise.
- Cost at scale
- A Standard index holding 100 GB and serving 30 million read units a month costs about $33 for storage plus $480 to $540 for reads, before writes and egress; the same traffic on Enterprise reads at $720 to $810.
- Limits
- 100 data-plane requests per second per namespace; 10,000 values per $in filter; Starter is AWS us-east-1 only with 5 indexes; Starter and Builder reads are blocked when quotas or the egress allowance run out.
What's good
- No infrastructure to run, tune or upgrade
- Full-text, vector, reranking and hosted embeddings in one API
- Builder plan caps spend at $20 a month with hard quotas
- BYOC on AWS, Azure and Google Cloud for Enterprise
- SSO and RBAC from the $50 Standard tier
What's not
- Reads cost $16 to $27 per million units, so busy indexes get expensive
- No open-source edition or local emulator since Pinecone Local was retired
- Private endpoints, CMEK, audit logs and uptime SLA are Enterprise-only
- Builder and Starter stop serving reads when quotas run out
- Most performance figures are self-reported
Best for
Teams without database operations staff who want managed vector search and accept usage-based fees.
Alternatives
Pinecone: common questions
- Is Pinecone worth it?
- Pinecone removes almost all the operations work from vector search and charges for it on every read, which suits a small team far better than a busy production index. We rate it 4.2 out of 5.
- How much does Pinecone cost?
- Free Starter (2 GB, 1M read units/month, AWS us-east-1 only). Builder $20/month flat with hard quotas. Standard $50/month minimum, then $0.33/GB/month storage, $16–$18 per 1M read units, $4–$4.50 per 1M write units. Enterprise $500/month minimum.
- What are the downsides of Pinecone?
- Reads cost $16 to $27 per million units, so busy indexes get expensive; no open-source edition or local emulator since Pinecone Local was retired; private endpoints, CMEK, audit logs and uptime SLA are Enterprise-only; builder and Starter stop serving reads when quotas run out; most performance figures are self-reported.
- Who is Pinecone best for?
- Teams without database operations staff who want managed vector search and accept usage-based fees.
- How much does Pinecone cost for a team or in production?
- A Standard index holding 100 GB and serving 30 million read units a month costs about $33 for storage plus $480 to $540 for reads, before writes and egress; the same traffic on Enterprise reads at $720 to $810.
- Is Pinecone safe for company data?
- SOC 2 Type II, ISO 27001 and GDPR on paid tiers; HIPAA as a $190-a-month add-on on Standard or included on Enterprise; data encrypted with AES-256 at rest; BYOC runs inside your own cloud account on Enterprise.
- What are the best Pinecone alternatives?
- The closest alternatives we have reviewed are turbopuffer.
Answers checked Oct 23, 2026.
Make Pinecone? Show this rating on your site
