How much does one AI model API call cost?
Estimate input and output charges with sourced rates and clearly stated usage assumptions.
API pricing often states a rate per million tokens, while your bill accumulates across individual requests. Confirm the model, platform and mode, then multiply each billable usage item by its matching rate. A unit price is not the total request price.
General calculation
For a rate per million tokens:
Item cost = billable tokens ÷ 1,000,000 × rate per million tokens
Calculate input, output and any other billed items separately, then add their costs in one currency. For video billed per second, multiply billable seconds by the per-second rate for the selected resolution.
Worked example: one text request
The GPT-6.1 Sol provider page listed standard text rates of USD 2 per million input tokens and USD 10 per million output tokens when checked on 2026-10-31. Its API ID is gpt-6.1-sol; see our model profile.
These are hypothetical usage numbers, not a real API call: one Standard request uses 2,000 ordinary uncached text input tokens and 1,000 billed output tokens. It has no tools, image input, regional processing or other charged features, and remains below the provider’s long-context premium threshold. Taxes and currency conversion are excluded.
| Item | Usage and rate | Calculation | Subtotal |
|---|---|---|---|
| Text input | 2,000 tokens; USD 2 per million | 2,000 ÷ 1,000,000 × 2 | USD 0.004 |
| Text output | 1,000 tokens; USD 10 per million | 1,000 ÷ 1,000,000 × 10 | USD 0.010 |
| Total | These two items | 0.004 + 0.010 | USD 0.014 |
If 100 requests have exactly the same usage and conditions, the estimate is 100 × USD 0.014 = USD 1.40. Real token counts vary. Check actual usage records and provider bills rather than treating this as a fixed per-call price.
Why the bill may differ
- Conversation history, attached files and output length change usage.
- Cached input, cache writes, Batch, Flex and Fast modes may have different rates. Use the rate for the mode actually selected.
- Tools, image input, long context and regional processing may add charges not included in this example.
- Image, video and audio tasks have different units; some combine input and output charges. Check our price table and the official billing rules.
Record model ID, platform, mode, currency, source, check date and every usage quantity. If an item is missing, label the result a partial cost estimate. See how to read API prices and how to verify sources. Recheck the provider page for current terms.