See what an AI feature really costs — per request, per day, and per month — before you ship it.
Estimate your LLM API spend
Enter average tokens per request and daily volume to see per-request, per-day, and per-month costs for every model, cheapest first. Prices as listed by providers, Oct 2026 — verify before budgeting.
Prices as listed by providers, Oct 2026 — verify before budgeting. Providers change prices and run discounts, so treat every figure here as an estimate for planning, not a quote.
| Model | Provider | Per request | Per day | Per month |
|---|
Monthly figures are the daily cost multiplied by 30. Token counts are averages — long prompts, system instructions, and retries all add tokens, so measure your real usage before committing to a budget.
About this tool
Every AI feature looks cheap in a demo and expensive in production, because the real driver is not the price on a pricing page — it is tokens multiplied by volume. This calculator turns published per-million-token list prices into the three numbers you actually budget with: cost per request, cost per day, and cost per month. Enter the average input tokens in one request (your prompt, system instructions, and any retrieved context), the average output tokens the model returns, and how many requests you expect per day. The selected model's cost appears instantly, and the comparison table re-prices the exact same workload on every model in the list, sorted cheapest first, so you can see what switching models would save — or cost — at your real volume. It is built for the questions teams ask before launch: What does one support chat cost us? What happens at ten thousand requests a day? Is the stronger model worth several times the price for this feature? All figures are estimates based on the listed prices shown on this page; providers discount, reprice, and bill for extras like caching and retries, so verify current prices with the provider before you commit to a budget.
How it works
- Pick a model from the list — the comparison table always shows every model for the same workload.
- Enter your average input tokens per request, counting the system prompt and any context you send, not just the user's message.
- Enter your average output tokens per request and your expected requests per day.
- Read the per-request, per-day, and per-month (30-day) costs, and compare models in the table, which sorts cheapest first and updates as you type.
Frequently asked questions
How is the monthly cost calculated?
Per-request cost is (input tokens / 1,000,000 × input price) + (output tokens / 1,000,000 × output price). That is multiplied by requests per day for the daily figure, and by 30 for the monthly estimate.
Why do input and output tokens have different prices?
Providers price them separately, and output tokens are usually several times more expensive than input tokens. A workload with short prompts and long answers can cost far more than the input price alone suggests.
Are these prices guaranteed?
No. The table uses prices as listed by providers, Oct 2026 — verify before budgeting. Providers change prices, offer batch and caching discounts, and may bill differently for features like long context, so treat the results as planning estimates.
What counts toward input tokens?
Everything you send: the system prompt, conversation history, retrieved documents, tool definitions, and the user's message. In chat applications the history grows with every turn, so measure a real full conversation, not just the first message.
Which token counts should I enter if I am not sure?
Start with a rough estimate, then log real usage from a test run and re-enter the measured averages. Even a rough volume estimate is useful, because the comparison between models matters more than false precision in the totals.