- Set the workload
Enter documents and any other volume assumptions that define the period or batch.
- Add usage assumptions
Complete the remaining usage fields, keeping token, duration, or size units consistent with their labels.
- Enter current rates
Use the current prices for the exact model, endpoint, quality, and processing tier you plan to use.
- Review the estimate
Read the main result and the supporting breakdown to see the largest cost or capacity driver.
- Test another scenario
Change one assumption at a time or select Reset to restore the default example values.
OpenAI Embedding Calculator
The OpenAI Embedding Calculator estimates embedding token volume, API cost, and vector-storage size. It supports document counts, average tokens per document, chunk overlap, vector dimensions, and a chosen price per million tokens.
Teams can use the result to budget an initial indexing run or recurring ingestion pipeline. Storage is an approximation because database metadata, indexes, compression, and replication can add substantially to the raw vector bytes.
Calculation inputs
documents
tokens
%
USD
dimensions
bytes
Result
—
Estimated result
Billable tokens—
Raw vector storage—
Vectors—
Effective tokens/document—
Billable tokens = Documents × Average tokens × (1 + overlap percentage ÷ 100). API cost = Billable tokens × price per 1M ÷ 1,000,000. Raw vector bytes = Documents × dimensions × bytes per dimension.
The calculator applies the entered values directly and displays rounded results for planning.
What the result means
The main result estimates embedding generation cost; the breakdown also shows raw vector storage.
A document can produce multiple vectors when chunked, so replace the document count with the actual chunk count when known.
Given: Example assumptions
Calculation: For 100,000 documents averaging 800 tokens with 15% overlap, billable volume is 92,000,000 tokens. At $0.02 per million tokens, embedding cost is $1.84. With 1,536 dimensions stored as four-byte floats, raw vectors require 614,400,000 bytes, about 0.57 GiB.
Result: The displayed result follows the same formula and rounding logic.
Should I enter documents or chunks?
Use the number of vectors you expect to create. If each document becomes several chunks, enter total chunks for a better storage estimate.
What does overlap overhead mean?
It approximates duplicated tokens caused by overlapping chunk boundaries. Set it to zero when your measured token total already includes overlap.
Does the storage estimate include metadata?
No. It covers raw vector values only and excludes IDs, text, metadata, indexes, replication, and database overhead.
Which byte size should I use?
Four bytes represents float32 values. Use the byte width of your actual vector encoding or quantization method.
Are query embeddings included?
Only if you include them in the entered volume. Recurring query embeddings can be modeled as additional documents or a separate run.