See exactly where a single prompt's cost comes from — system prompt, user prompt, and output — then project daily, monthly, and annual spend.
Enter your details and click Calculate to see results
Most AI cost calculators treat a request as one lump of input tokens and one lump of output tokens, which hides where the money is actually going. A typical production prompt is really three distinct pieces stitched together: a fixed system prompt that repeats on every call, a variable user prompt that changes per request, and the model's generated output. This prompt cost estimator itemizes those three pieces separately so you can see, for example, that your system prompt alone might be costing more per month than all of your user input combined — a common and easily fixable problem.
Paste your system prompt and a representative user prompt into the two text boxes; the calculator estimates their token counts from character length, then multiplies each by your selected model's input token price to get a system prompt cost and a user prompt cost. It adds those for a total input cost, computes output cost from your expected output tokens at the model's output price, and sums everything into a cost per prompt. That per-prompt figure is then projected out to daily, monthly, and annual totals based on how many prompts you run per day, with a doughnut chart showing what share of the total each of the three components represents.
Because system prompts are static and repeat on every single call, they are usually the easiest cost lever to pull — trimming an unnecessarily verbose system prompt, or enabling prompt caching where a provider supports it, can cut a meaningful percentage off your total bill without touching the user-facing experience at all. Seeing the breakdown also helps you decide where caching would actually help: if your user prompt cost dominates (highly variable input, like large pasted documents), caching the system prompt alone won't move the needle much.
Your fixed instructions, persona, and formatting rules, billed at the model's input price on every single request. Because it repeats identically each call, it's the best candidate for prompt caching discounts where the provider offers them.
The variable part of the request — the actual question, document, or task from the end user — also billed at the input price, but its length changes call to call, so its total cost scales with real usage patterns rather than being fixed.
The model's generated response, billed at the (usually much higher) output token price. Because this rate is 3-5x the input rate on most providers, output length is frequently the single biggest lever for reducing per-prompt cost.
Common questions about prompt cost breakdowns
Explore other AI & tech tools