- Set the workload
Enter prompt calls 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 Prompt Calculator
The OpenAI Prompt Calculator estimates the token footprint and recurring cost of a reusable prompt. It combines system instructions, user content, examples, retrieved context, and expected output length, then scales the result by the number of calls.
This is useful for prompt engineering reviews because it exposes which prompt component consumes the most context and budget. It can also show the effect of shortening examples or moving stable text into a cache-friendly prefix.
Calculation inputs
calls
tokens
tokens
tokens
tokens
tokens
USD
USD
Result
—
Estimated result
Input tokens per call—
Total tokens per call—
Cost per call—
Context share of input—
Input tokens per call = System + User + Examples + Retrieved context. Total cost = Calls × [(Input tokens × Input rate + Output tokens × Output rate) ÷ 1,000,000]
The calculator applies the entered values directly and displays rounded results for planning.
What the result means
The result estimates the recurring text-processing cost of the full prompt pattern.
Token estimates should be replaced with measured tokenizer or API usage data when available.
Given: Example assumptions
Calculation: Given 5,000 calls with 700 system, 300 user, 500 example, 1,000 context, and 350 output tokens, input is 2,500 tokens per call. At $2.50 input and $15 output per million, cost per call is (2,500 × 2.50 + 350 × 15) ÷ 1,000,000 = $0.0115, so the total is $57.50.
Result: The displayed result follows the same formula and rounding logic.
Do characters equal tokens?
No. Token counts depend on the tokenizer and language, so character counts are only a rough proxy.
Should examples be counted on every call?
Count them whenever they are sent. If a stable prefix receives cached pricing, model that separately with the Token Calculator.
What belongs in retrieved context?
Include text inserted by search, RAG, memory, or database lookups before the model generates an answer.
Does output length affect input context?
It does not increase input tokens, but it uses the model context window and is normally billed at the output rate.
How can I reduce prompt cost?
Remove redundant instructions, shorten examples, retrieve fewer higher-quality passages, and cap output length where appropriate.