OpenAI Cost Calculator
Project your monthly and annual OpenAI API spend from request volume and average token sizes.
Formula last reviewed 4 August 2026 · How we verify our calculators
Projected monthly cost
$300.00
- Daily cost
- $10.00
- Annual cost
- $3650.00
- Cost per request
- $0.01000
Updates live as you type
Frequently asked questions
Multiply per-request cost by requests per day, then by 30 for the month. Per-request cost is (input tokens × input rate + output tokens × output rate) per million tokens.
Caching reuses a static prompt prefix at a steep discount (commonly ~90% off cached input tokens), which is significant for apps that send the same long system prompt on every call.
Real traffic varies per request, so averages give a reliable monthly projection. Measure a sample of real requests to set accurate input/output averages.
The ChatGPT token-cost tool estimates tokens from pasted text for one-off checks; this one projects ongoing spend from your request volumes.
A one-cent request, times a thousand a day, times a year
GPT-5.4, 1,000 requests a day, averaging 1,000 input and 500 output tokens per request: per-request cost works out to (1,000 ÷ 1,000,000 × $2.50) + (500 ÷ 1,000,000 × $15) = $0.0025 + $0.0075 = $0.01. That single cent, multiplied across 1,000 requests a day, becomes a daily cost of $10, projecting to $300 a month and $3,650 a year — a number easy to lose sight of when you're only looking at the per-request figure.
Why the input-output split matters more than the total token count
Output tokens typically cost several times more than input tokens, so the *mix* of a workload matters as much as its volume: chat applications with short prompts but long generated replies are output-heavy, while classification or extraction jobs that return a short label are input-heavy. In the example above, output cost is 75% of the per-request total — which is also why prompt caching, which only discounts cached *input* tokens by roughly 90%, has a smaller effect here than it would on a more input-heavy workload.
Why guessed averages produce unreliable projections
This calculator's accuracy depends entirely on the input and output token averages you feed it — measuring a representative sample of real requests, rather than guessing round numbers, is what turns this from a rough estimate into a reliable monthly budget. Logging actual input/output token counts from a week of production traffic gives a materially better basis than assuming your averages match the defaults.
Volume forecasting versus a one-off check
This tool is built for volume-based forecasting — projecting ongoing spend from request patterns over time. To estimate tokens from one specific piece of text before sending it, the ChatGPT token-cost calculator is the better fit instead.
Pricing as of June 2026. LLM rates change frequently — verify current prices on the provider's official pricing page before budgeting.