Pinecone
Pinecone pricing meters four things on a serverless index: read units, write units, gigabytes stored and gigabytes returned.
Updated on:
Pinecone pricing: three meters, and reads cost four times writes
Pinecone pricing meters four things on a serverless index: read units, write units, gigabytes stored and gigabytes returned. Reads and writes carry different prices, and the gap is exact. A read unit costs four times a write unit in every published cell of the rate card, and Pinecone documents the query-to-RU conversion precisely.
Key takeaways
A query costs 1 read unit per 1 GB of namespace size, with a 0.25 RU floor. top_k, include_metadata and include_values don't change RU cost; only namespace size does.
Reads cost exactly 4x writes: $16 per million RU against $4 per million WU on Standard, $24 against $6 on Enterprise. The ratio holds across every regional band.
Enterprise pays 50% more per read and write unit than Standard at the same $0.33/GB storage rate. The higher tier buys SLA, BYOC and compliance, not cheaper units.
Starter and Builder block at quota with a 429 instead of billing an overage. Only Standard and Enterprise turn excess usage into an invoice.
Pinecone pricing in 2026
Tier | Price | What's included | Metered limit
|
|---|---|---|---|
Starter | Free | 1 project, 5 indexes, us-east-1 only | 1M RU, 2M WU, 2 GB, 1 GB egress |
Builder | $20/month flat | All GA regions, 10 indexes per project | 2M RU, 5M WU, 10 GB, 10 GB egress. No overages |
Standard | $50/month min. | Dedicated read nodes, import, backups, RBAC, SSO | $16-$18/M RU, $4-$4.50/M WU, $0.33/GB/mo, 100 GB egress then $0.10/GB |
Enterprise | $500/month min. | 99.95% SLA, BYOC, private endpoints, CMEK, HIPAA | $24-$27/M RU, $6-$6.75/M WU, $0.33/GB/mo, 100 GB egress then $0.10/GB |
What Pinecone pricing actually meters
Four serverless meters, and the read conversion is the one worth knowing.
Read units. A query uses 1 RU per 1 GB of namespace size, minimum 0.25 RU. Response shape is irrelevant: Pinecone states that top_k, include_metadata and include_values don't affect query cost, since only the scanned namespace sets it. A 10 GB namespace costs 10 RU per query, or $0.00016 at the $16 Standard rate. Fetch bills 1 RU per 10 records with a 1 RU floor, and list is flat at 1 RU per call covering up to 100 records.
Write units. An upsert uses 1 WU per 1 KB of request, minimum 5 WU. Updates charge for the new record and the existing one, so a 6.24 KB record replacing a 6.50 KB one costs 13 WU. Deletes bill at 1 WU per KB removed, and dropping a namespace costs a flat 5 WU.
Storage and egress. Storage runs $0.33/GB/month on both paid tiers. Egress became its own meter on 1 September 2026, at $0.10/GB past the allowance.
The asymmetry looks deliberate. Reads cost more because a query scans the namespace, so cost tracks index size rather than result size. Namespace design, not query tuning, is the lever on the bill.
How credits and minimums work
Pinecone runs prepaid credits and monthly minimums as two separate mechanics.
Prepaid credits come in a band: buy between $8,000 and $25,000 upfront and earn discounted rates plus extra capacity that varies by plan and amount. Credits apply at list price, and anything past them bills at list price too. Expiry and rollover rules aren't documented; the one credit Pinecone did date, a $250 bulk import grant, expired 60 days after issue.
The monthly minimum is a floor you grow into, not a fee on top. A Standard account using $20 in August gets a $50 invoice: $20 of line items, plus one $30 line covering the rest of the commitment. Cross $50 and that line disappears. Builder differs: its $20 is a flat fee, and usage past Builder limits gets blocked rather than billed.
What happens when you hit the limit
It depends on which half of the lineup you're on.
Flat-fee plans stop. Starter and Builder return 429 - TOO_MANY_REQUESTS once you exhaust monthly read units, write units or embedding tokens. Egress behaves the same way: in-scope reads throw RESOURCE_EXHAUSTED while index statistics and writes keep working. Allowances reset next period.
Usage-based plans keep serving and bill the difference, since Standard and Enterprise carry no monthly RU, WU or token ceiling. Throughput limits still apply to everyone: 2,000 query read units per second per index and 100 query requests per second per namespace, identical on all four plans. Under non-payment, requests return 402 - Payment Required and a deactivated account keeps data for 30 days before deletion.
How Pinecone pricing has changed across all these years
Date | Milestone | Source
|
|---|---|---|
1 Sep 2026 | Egress metering starts, per-plan allowances, $0.10/GB beyond | Vendor |
1 Sep 2026 | $250 bulk import credit withdrawn for new subscriptions | Vendor |
1 Jun 2026 | Bulk import overage cut to $0.25/GB from $1/GB | Vendor |
4 May 2026 | Builder launches at $20/month flat with no overages | Vendor |
1 Apr 2026 | Assistant goes fully usage-based, hourly per-assistant fee removed | Vendor |
1 Feb 2026 | HIPAA add-on introduced for Standard at $190/month | Vendor |
18 Aug 2025 | Pod-based indexes closed to new Standard and Enterprise sign-ups | Vendor |
Customer
Sentiment Highlights
"Pricing can increase quickly as the number of vectors and queries grows."
Vamshi K., AI software engineer, mid-market, G2, September 2026
"understanding how different usage levels, index configurations, and data volumes affect cost"
Raphael J., Lead AI Engineer, on the main challenge, G2, September 2026
Explore other providers
ChatGPT
Chatbot
ChatGPT pricing sells seats from $0 to $100 a month and meters the ambitious work on top in credits.
n8n
Developer Tool
n8n pricing charges per workflow execution, and an execution is one run of an entire workflow no matter how many steps or how much data it processes.
Perplexity
Chatbot
Perplexity pricing sells six seats and counts usage in two units.
How much does Pinecone cost per query?
What is a read unit in Pinecone?
Does Pinecone have a free tier?
Can you still buy pod-based indexes?
