WORKLOAD ESTIMATE

What will your LLM workload cost?

Choose a model or enter your rates. Add the tokens for your workload to estimate the API cost and compare it with the matching official price.

Selected model rates from 0.3× official API pricing

See model prices Create an account

Calculate a workload.

Enter tokens for one representative request, then set how many similar requests you will run. The calculator selects the context tier before multiplying the cost. Estimate requests with different context lengths separately.

MODEL COST ESTIMATEUSD · rates per 1M tokens

Model Account group Number of identical requests

Billing categoryTokens per requestCodimixa rateOfficial rate

Uncached input

Output

Include cache categories

Enter tokens for one request. The calculator selects its context tier, then multiplies by the number of identical requests. Count each input token once.

Choose a model and enter tokens per request.

API token charges only. Add hosting, subscriptions, paid tools and explicit cache-storage fees separately.

Use separate billing categories.

Multiply the tokens in each category by its price per million, then add the amounts. Use uncached input separately from any billed cache reads or writes.

text

estimated API cost = Σ (tokens in category ÷ 1,000,000 × category rate)

Turn a task sample into a monthly estimate.

Measure a representative set of completed tasks. Divide its API cost by completed tasks, then multiply by the number you expect to run. Include failed attempts in the sample cost.

Keep the rest of your budget visible.

This calculator estimates model API charges. Add hosting, software plans and paid external tools separately. A change in context length, retries or the selected model can change your estimate.

Reduce repeated API spend

A FEW DETAILS

Before you continue

Is this a quote?

No. It is an estimate from the rates and token counts you supply. Your account’s applicable rates and actual billed usage determine the charge.

Should I multiply the total by retries?

Only if your token totals exclude retry attempts. If they already include every request, another retry multiplier would count the same work twice.