Vector buckets use pay-as-you-go billing. Three billing items apply: storage fees based on the amount of vector data stored, data retrieval fees based on the amount of data scanned per query, and API request fees based on the number of API calls.
Vector buckets are currently in invitational preview and are free of charge. The billing information below is for reference only. Final billing rules will be announced at the commercial launch.
Storage fees
| Billable item | Billing condition | Billing rule | Billing cycle | Billing method |
|---|---|---|---|---|
| Storage fees | Storing vector data in a vector bucket | Billed based on the size of data written | Hourly | Pay-as-you-go (USD/GB/month) |
Billing rules
Total storage is the sum of three components:
Total storage size = Size of vector primary keys + Size of vector data + Size of metadata
| Component | Calculation |
|---|---|
| Size of vector primary keys | Total size of all vector primary key strings (UTF-8 encoding) |
| Size of vector data | (Vector dimensions ÷ 1024) × 4 bytes × Number of vectors |
| Size of metadata | Total size of all metadata key and value strings (UTF-8 encoding) |
Example: calculating storage size
Assume a vector bucket for document retrieval with the following configuration:
| Parameter | Value |
|---|---|
| Vector dimensions | 1024 |
| Number of vectors | 100,000 |
| Data type | float32 (4 bytes per dimension) |
| Average vector primary key length | 20 bytes |
| Average metadata size | 200 bytes/vector |
Calculation:
Size of vector primary keys:
100,000 × 20 bytes = 2,000,000 bytes ≈ 1.91 MBSize of vector data:
(1024 ÷ 1024) × 4 bytes × 100,000 = 400,000 bytes ≈ 0.38 MBSize of metadata:
100,000 × 200 bytes = 20,000,000 bytes ≈ 19.07 MB
Total storage size = 1.91 MB + 0.38 MB + 19.07 MB = 21.36 MB
Quick reference: storage size by scale
The following table shows estimated vector data sizes for common configurations. Primary key size and metadata size are not included and vary by use case.
| Vectors | Dimensions | Data type | Vector data size |
|---|---|---|---|
| 100,000 | 128 | float32 | 0.05 MB |
| 100,000 | 768 | float32 | 0.29 MB |
| 100,000 | 1024 | float32 | 0.38 MB |
| 1,000,000 | 768 | float32 | 2.86 MB |
| 1,000,000 | 1536 | float32 | 5.72 MB |
The values above cover vector data only. Add the size of your primary keys and metadata to get the total billable storage size.
Data retrieval fees
| Billable item | Billing condition | Billing rule | Billing cycle | Billing method |
|---|---|---|---|---|
| Data retrieval fees | Initiating a query to retrieve vector data | Billed based on the index data size for each retrieval operation | Hourly | Pay-as-you-go (USD/TB) |
Billing rules
Each vector search query is billed based on the total number of vectors in the index, regardless of how many results are returned.
Retrieved data size per query = Number of vectors in the index × (Size of vector primary key + Size of vector data + Size of filterable metadata)
Where Size of vector data per vector = Vector dimensions ÷ 1024 × 4 bytes
The number of vectors in the index is the total count of indexed vectors, not the number of results returned by a query. Parameters such as top_k do not affect the retrieval fee.Example: calculating retrieved data size
Assume a vector search with the following parameters:
| Parameter | Value |
|---|---|
| Vector dimensions | 1024 |
| Number of vectors in the index | 100,000 |
| Average vector primary key length | 20 bytes |
| Average filterable metadata size | 100 bytes/vector |
Calculation:
Size per row:
20 bytes + (1024 ÷ 1024 × 4 bytes) + 100 bytes = 124 bytesTotal retrieved data size:
100,000 × 124 bytes = 12,400,000 bytes ≈ 11.83 MB
Quick reference: retrieved data size per query
The following table shows estimated retrieved data sizes for common index configurations, assuming 20-byte primary keys and 100-byte filterable metadata per vector.
| Vectors in index | Dimensions | Size per row | Size per query |
|---|---|---|---|
| 100,000 | 128 | 120.5 bytes | 11.49 MB |
| 100,000 | 768 | 123 bytes | 11.73 MB |
| 100,000 | 1024 | 124 bytes | 11.83 MB |
| 1,000,000 | 768 | 123 bytes | 117.3 MB |
| 1,000,000 | 1536 | 126 bytes | 120.2 MB |
Retrieved data size per query is dominated by primary key size and filterable metadata size. Vector dimensions have a smaller effect because the formula normalizes to a 1024-dimension baseline.
API request fees
API requests are classified as PUT or GET requests:
| Request type | Operations |
|---|---|
| PUT requests | PutVectorBucket, PutVectorIndex, PutVectors, DeleteVectorBucket, DeleteVectorIndex, DeleteVectors, ListVectorBuckets, ListVectorIndexes, ListVectors |
| GET requests | GetVectorBucket, GetVectorIndex, GetVectors, QueryVectors |