LLM API Cost Calculator

Example#

With an input price of $2.00 and an output price of $8.00 per million tokens, 5,000 requests per day, 1,200 input tokens and 350 output tokens per request, and 30 days in a month, the estimate is about $0.0052 per request, $26.00 per day, $780.00 per month and $9,360.00 per year, using 232.5 million tokens a month.

If half of the input tokens hit a cache priced at $0.50 per million, the same usage falls to about $0.0043 per request and $645.00 per month.

How it works#

cost per request = ( input tokens × (1 − hit rate) × input price
                  + input tokens × hit rate × cached price
                  + output tokens × output price ) ÷ 1,000,000
monthly cost     = cost per request × requests per month
yearly cost      = monthly cost × 12

Prices are in US dollars per million tokens. When you enter requests per day, the tool multiplies by the days per month you set. Calculations use full precision and only the display is rounded: money to the cent, and the per-request cost to four decimals. The share bar shows how much of the cost comes from input and how much from output tokens.

If model prices have been added to the site, you can pick a model and the tool compares every model on the same usage, sorted by monthly cost, with the date each price was checked and a link to the provider's official page. You can always type your own prices instead.

Assumptions#

• Every request uses the average token counts you enter.
• Prices are per million tokens in USD, taken from the provider's official pricing page on the date shown.
• Reasoning models may bill internal reasoning tokens as output tokens; include them in your output token count.

Use cases#

• Budgeting a new AI feature before you build it.
• Comparing the monthly cost of two models for the same workload.
• Seeing how much prompt caching could save.
• Explaining expected spend to a team or a client.

Limitations#

This is an estimate, not an invoice. It leaves out taxes, batch or committed-use discounts, free tiers, long-context price tiers, image and audio tokens, tool-call overhead and failed requests. Providers change prices, so check the date next to each price and the provider's page before you commit. The calculation runs in your browser and nothing you enter is sent anywhere.