Pinecone Pricing 2026: Every Plan, Rate and Real Cost

Last Updated: November 15, 2026

Summary: Pinecone has four plans in 2026: free Starter, a new Builder plan at a flat $20 a month, Standard at a $50 monthly minimum and Enterprise at a $500 monthly minimum, plus custom BYOC (bring your own cloud) deals. On Standard, you pay $0.33 per GB per month for storage, $16 to $18 per million read units and $4 to $4.50 per million write units, depending on cloud and region. Small apps often cost $0 to $50 a month, but queries against large namespaces can push bills into the thousands.

Pinecone pricing is easy to misread because the cost of a query depends on how much data it searches, not how many results it returns. This guide covers every plan and unit rate, three worked cost examples, ways to cut the bill, and how Pinecone compares with Weaviate, Qdrant, Zilliz, MongoDB Atlas and pgvector. Prices were checked against Pinecone's official pricing page and documentation on November 12, 2026.

Pinecone costs $0 on the Starter plan, $20 a month flat on Builder, at least $50 a month on Standard and at least $500 a month on Enterprise. On paid plans the minimum is a floor, not an extra fee: if usage comes to less, Pinecone bills the difference.

Plans as listed on Pinecone's pricing page and in its object limits documentation:

Pinecone launched Builder on May 4, 2026, according to its 2026 release notes, and opened it to all generally available regions in June. Builder never bills overages: at a quota, Pinecone blocks the operation instead.

Yes. Pinecone's Starter plan is free, and Pinecone says many small workloads fit inside it. The catch is the caps: 2 GB of storage, 1 million read units and 2 million write units a month, all in AWS us-east-1.

Starter also includes 5 million embedding tokens and 500 reranking requests a month.

To test Standard first, Pinecone offers a 21-day Standard trial with $300 in credits and no up-front payment, according to its trial documentation. When it ends, quotas are blocked until you add a payment method or change plans, and downgrading to Starter deletes anything beyond Starter's limits, including backups.

Pinecone's usage pricing has three main meters: storage, read units (RUs) and write units (WUs). Enterprise pays 50% more per read and write unit in exchange for its SLA, security features and included support.

Since October 2026, restoring a backup into a different region also adds a per-GB cross-region transfer fee, for example $0.023 per GB between most US and EU AWS regions.

Read units scale with the size of the namespace you search, not the number of results. According to Pinecone's cost documentation, the rules are:

Storage is calculated per record as ID bytes plus metadata bytes plus 4 bytes per vector dimension. Pinecone's example: 500,000 records with 8-byte IDs, 768 dimensions and 500 bytes of metadata take about 1.79 GB.

Pinecone also sells hosted embedding and reranking models, a managed RAG product called Pinecone Assistant, and paid support, all billed separately from the database.

Inference (embeddings and reranking):

Pinecone Assistant: on Standard and Enterprise, Assistant costs $3 per GB per month of file storage, $8 per million input tokens, $15 per million output tokens, $5 per million context tokens and $0.0005 per ingestion unit, per Pinecone's Assistant documentation, with no hourly fee.

Some third-party guides still list Pro support at $499 a month and an hourly Assistant fee. Both are out of date.

The math below uses Pinecone's published formulas and the low end of the Standard rate ranges ($16 per million RUs, $4 per million WUs), with the high end in brackets. Assumptions: 16-byte record IDs, sizes in decimal GB, fractional read units billed proportionally, and inference costs excluded unless stated.

A company indexes 100,000 document chunks (1,536 dimensions, 1 KB of metadata) and answers 20,000 employee questions a month.

Monthly cost: $0. Every meter sits well under the Starter caps. If the company needs a cloud or region other than AWS us-east-1, Builder covers the same workload for $20.

A help-center search covers 1 million records (1,024 dimensions, 500 bytes of metadata), serves 300,000 searches a month and updates 50,000 records.

Usage total: $25.50 ($28.50). On Standard, that is below the $50 minimum, so the bill is $50. On Builder, it fits the included allowances for a flat $20 (egress is about 1.5 GB). One warning: the initial load of 1 million records uses about 4.6 million WUs, so first-month writes plus updates would exceed Builder's 5M cap. Spread the load or use the Standard trial.

A SaaS company stores 20 million records (768 dimensions, 300 bytes of metadata) for 1,000 customers, runs 10 million queries a month and adds 2 million new records a month.

Monthly cost: about $96.24 ($104.63). Embedding 10 million 20-token queries with llama-text-embed-v2 would add $32. The contrast: in a single namespace, every query would scan 67.76 GB and bill 67.76 RUs, so 10 million queries would cost $10,841.60 ($12,196.80) in reads alone. Same data, same traffic. Only the namespace design changed. If a few large customers hold most of the data, their namespaces cost more per query, so the real gap would be smaller, but still large.

The biggest savings come from shrinking what each query scans and each write sends. Pinecone's guide to saving on costs covers most of these steps:

Dedicated Read Nodes (DRN) replace per-query read units with a fixed hourly rate per node, which pays off for sustained, high query volume. DRN became generally available on Standard and Enterprise on April 15, 2026. Pinecone's DRN documentation gives one example rate: a b1 node in AWS us-east-1 at $336.42 a month ($0.46 an hour), so an index with two shards and two replicas costs $1,345.68 a month before storage, writes and egress.

In its DRN launch post, Pinecone reported cost reductions of 77% to 97% on three anonymized production workloads. With low or spiky traffic, you pay for idle nodes, so on-demand stays cheaper.

Pinecone bills purely by usage, with no servers to size. Most alternatives' paid tiers bill by cluster resources or by vector dimensions instead. Entry pricing from each vendor's official page, checked November 12, 2026:

Weaviate's pricing page lists Flex from $0.00465 per million vector dimensions and $0.12 per GiB of storage, with Premium priced through sales. Qdrant publishes no starting price for paid clusters. Zilliz's pricing page covers the managed version of open-source Milvus and lists Enterprise dedicated clusters from $197 a month.

MongoDB's documentation confirms Vector Search runs on Atlas Flex clusters ($8 to $30 a month, 5 GB of storage), with dedicated M10 clusters from $60 a month. pgvector is a free Postgres extension, so its cost is your Postgres bill: Supabase Pro starts at $25 a month with 8 GB of disk, plus $0.125 per extra GB.

Pick Pinecone if you want zero infrastructure work and traffic is low or spiky; Builder at $20 suits small production apps. Pick pgvector or MongoDB Atlas if your data already lives in Postgres or MongoDB and your vector counts are modest. Look at Qdrant, Weaviate or Zilliz dedicated clusters if you run heavy, steady query volume over one large dataset, where per-query read units get expensive, and price them against Pinecone's Dedicated Read Nodes.

A vector database stores embeddings, the lists of numbers AI models use to represent the meaning of text, images or other data, and finds the closest matches to a query. Our guide to what embeddings are explains the input side.

Most businesses use one for retrieval-augmented generation (RAG), where a chatbot looks up company documents before answering. Model calls are billed separately, so compare token costs in our OpenAI API pricing and Gemini API pricing guides before setting a budget.

How much does Pinecone cost per month?
Pinecone costs $0 on Starter, a flat $20 on Builder, at least $50 on Standard and at least $500 on Enterprise. On Standard, usage is billed at $0.33 per GB of storage, $16 to $18 per million read units and $4 to $4.50 per million write units.

Is Pinecone free to use?
Yes, Pinecone's Starter plan is free and includes 2 GB of storage, 1 million read units and 2 million write units a month. Starter indexes must run in AWS us-east-1, and reads are blocked once the 1 GB monthly egress allowance is used up.

What is a read unit in Pinecone?
A read unit is Pinecone's billing measure for queries, fetches and lists. A query costs 1 read unit per GB of the namespace it searches, with a minimum of 0.25, so querying a 10 GB namespace costs 10 read units.

What is the Pinecone Builder plan?
The Builder plan is a $20-a-month flat plan Pinecone launched in May 2026. It includes 10 GB of storage, 2 million read units, 5 million write units and 10 GB of egress a month on AWS, GCP or Azure, and it blocks usage at the limits instead of billing overages.

Is Pinecone cheaper than pgvector?
For small, low-traffic workloads, Pinecone's free Starter or $20 Builder plan can cost less than a paid managed Postgres plan such as Supabase Pro at $25 a month. At heavy query volume, pgvector on existing Postgres can be cheaper because you pay for compute, not per query.

Pinecone pricing in 2026 is cheap at the edges and tricky in the middle: prototypes run free on Starter, small production apps fit the $20 Builder plan, and Standard's $50 minimum only makes sense once you outgrow Builder's 10 GB and monthly unit caps.

The number that decides your bill is namespace size. Because every query bills read units based on the data it scans, splitting data into namespaces can turn a five-figure monthly bill into one closer to $100. Recheck Pinecone's pricing page before committing.

⭐ Get more AI Business Weekly in Google: add us as a preferred source on Google.

Written by Sameer Khan with AI assistance, fact-checked against the sources linked above.