Pinecone vs Weaviate vs Qdrant Cloud: Real Vector Database Cost for RAG in 2026
Pinecone rates are USD list prices for the Standard plan as reported by 2026 trackers; Pinecone's documentation defines the units. Weaviate rates are its published Flex and Premium prices. Qdrant Cloud does not publish per-unit prices, so it is sized, not priced.
The short answer
A small application costs nothing on Pinecone Starter or Qdrant's free cluster and $45 a month on Weaviate Flex. A growing application (2 million vectors, 1.5 million queries a month) costs about $355 a month on Pinecone if every query scans one 14 GB namespace, because Pinecone's documentation states that a query costs 1 read unit per GB of namespace size, minimum 0.25. Confining queries to small namespaces brings it to the $50 plan minimum. A high-volume single-namespace index (358 GB, 30 million queries) is calculated at about $172,000 a month under that formula, which is an argument for partitioning or Dedicated Read Nodes, not a realistic bill.
How each vendor meters
| Pinecone | Weaviate Cloud | Qdrant Cloud | |
|---|---|---|---|
| Plans | Starter (free: 2 GB, 2M write units, 1M read units), Builder ($20), Standard ($50 minimum), Enterprise ($500 minimum) | Free sandbox; Flex from $45/month; Premium from $400/month | Free forever: 0.5 vCPU, 1 GB RAM, about 4 GB disk; paid clusters billed hourly by resources |
| Reads | Read units at $16/M (Standard), $24/M (Enterprise). Query = 1 RU per GB of namespace size, minimum 0.25 RU; top_k and metadata settings do not change it; fetch 1 RU per 10 records | No per-query fee | No per-query fee |
| Writes | Write units at $4/M; 1 WU per KB of request, minimum 5 per request | Included in dimensions | Included in cluster |
| Storage | $0.33/GB/month | $0.12/GiB (Flex), $0.10/GiB (Premium) plus vector dimensions: from $0.00465/M (Flex), $0.003875/M (Premium shared) | Reserved disk |
| Backups | Not itemized in the rates reviewed | From $0.0264/GiB | Not itemized |
| Egress | $0.10/GB above 100 GB/month | Not itemized | Not itemized |
| Dedicated capacity | Dedicated Read Nodes: fixed hourly price, no RU limits | Premium dedicated deployments | Standard clusters |
Pinecone's documentation for read units changed: a 2023 page described RU growth as much slower than index size, while the current page states cost "scales linearly with the size of the targeted namespace"; this article follows the current page. One tracker lists Pinecone rates of $2/M writes and $8.25/M reads; the higher, more widely reported rates are used. The $16 per million read units and $4 per million write units are reported Standard-plan rates for Pinecone's baseline regions; Pinecone prices vary by cloud and region, no regional rate table was read, and other regions may cost more, so these are reported baseline rates, not confirmed for any one region. The $24 Enterprise read rate is reported the same way. Weaviate's "from" dimension rates apply to its most compressed index type; other index types list higher.
Workloads
Assumptions, illustrative: 1,536-dimension float32 embeddings (6,144 bytes) plus 1 KB of metadata, 7,168 bytes per vector; 10% of vectors rewritten each month; 10 results of about 1 KB returned per query.
| Small | Growing | High volume | |
|---|---|---|---|
| Vectors | 100,000 | 2,000,000 | 50,000,000 |
| Size | 0.72 GB | 14.3 GB | 358 GB |
| Queries per month | 30,000 | 1,500,000 | 30,000,000 |
Pinecone Standard, as billed
| Small | Growing | High volume | |
|---|---|---|---|
| Storage | $0.24 | $4.73 | $118.27 |
| Reads, one namespace | $0.34 | $344.06 | $172,032.00 |
| Writes | $0.29 | $5.73 | $143.36 |
| Egress | $0 | $0 | $20 |
| Usage | $0.87 | $354.52 | $172,313.63 |
| Billed | Starter $0 (within its allowances), or the $50 Standard minimum | $354.52 | $172,313.63 (illustrative single-namespace worst case) |
| Reads if queries hit 0.1 GB partitions | $0.12 | $6.00 | $120.00 |
| Billed with partitions | $0 / $50 | $50 (usage $16.46 is below the minimum) | $401.63 |
Formulas: reads = queries × max(0.25, namespace GB) × $16 ÷ 1,000,000; growing single namespace = 1,500,000 × 14.336 × $16 ÷ 1,000,000 = $344.06. Egress at high volume = 30,000,000 × 10 × 1 KB = 300 GB, less 100 GB included, × $0.10 = $20.
Weaviate Cloud (list arithmetic, one replica)
| Small | Growing | High volume (Premium shared) | |
|---|---|---|---|
| Vector dimensions | 153.6M | 3,072M | 76,800M |
| Dimension charge | $0.71 | $14.28 | $297.60 |
| Storage | $0.09 | $1.72 | $35.84 |
| Usage | $0.80 | $16.00 | $333.44 |
| Billed (minimum applies) | $45 | $45 | $400 |
Weaviate's published pricing does not state whether replicas multiply the dimension charge on a highly available cluster; if they do, a replication factor of 3 multiplies the high-volume dimension line to $892.80 and lifts that bill above the $400 minimum. Backups at the Flex rate on 358 GiB would add $9.45 a month.
Qdrant Cloud (sizing only)
| Small | Growing | High volume | |
|---|---|---|---|
| Raw vector memory (6,144 B × vectors) | 0.61 GB | 12.3 GB | 307 GB |
| With int8 quantization (÷4) | 0.15 GB | 3.1 GB | 76.8 GB |
| With binary quantization (÷32) | 0.02 GB | 0.4 GB | 9.6 GB |
| Cost | $0 on the free cluster (fits 1 GB RAM) | Not priced here | Not priced here |
Qdrant's pricing page directs buyers to an interactive calculator by cloud, region and node size; a cluster price for the growing and high-volume cases requires running it with the memory figures above.
What the advertised price misses
Namespace size drives Pinecone reads; the same 30 million queries cost $172,032 or $120 depending on partitioning. Dimensions scale storage and reads: a 384-dimension model (2,560 bytes per vector) cuts the growing workload's Pinecone usage to about $126 a month. Minimums decide small workloads: all three are under $1 in usage. Dedicated clusters bill while idle; Pinecone on-demand does not beyond its minimum.
Break-even
Pinecone on-demand reads cost about $16 per million read units, so each query costs about $0.000229 in a 14.3 GB namespace (14.336 RU × $16 ÷ 1,000,000). Including the growing workload's other Pinecone items (storage $4.73 and writes $5.73, $10.47 in total; egress is zero at that volume), Pinecone's usage equals Weaviate Flex's $45 minimum at about 151,000 queries a month and Weaviate Premium's $400 minimum at about 1.70 million queries a month. These crossovers apply only to this workload and a single unpartitioned namespace, exclude Weaviate's own replication and backup uncertainty, and would move with partitioning, dimensions and record size.
Sources
Pinecone: understanding cost, pricing; rates as reported by Orb, CostBench (Jul 2026) and UsagePricing (verified 11 Aug 2026). Weaviate: pricing, pricing update; unit rates as reported by Markaicode (3 Sep 2026). Qdrant: qdrant.tech/pricing.