Back to articles

Calculator guide

LLM API cost calculator examples for five common workloads

Token calculators are easiest to trust when the inputs start from a real workload shape. This guide illustrates five calculator presets, then routes each workload to a guide where its assumptions and limitations are easier to inspect.

Calculator examples Data as of catalog generation

Preset map

Start with the preset that matches the product behavior

Presets are input helpers, not model recommendations. Change the token counts and monthly request volume before making a production decision.

Workload Input tokens Output tokens Monthly requests Best next page
Chatbot
chatbot preset
1,000 500 10,000 Estimate cost · Open guide
RAG answer
RAG answer preset
4,000 800 5,000 Estimate cost · Open guide
Summarization
summarization preset
20,000 1,000 1,000 Estimate cost · Open guide
High-volume extraction
structured extraction preset
400 60 60,000 Estimate cost · Open guide
Coding agent
coding-agent preset
8,000 4,000 2,000 Estimate cost · Open guide

For a working budget, enter typical measured input and output tokens and monthly request volume in the calculator. Rerun with high-usage request measurements, actual cache behavior, retries, and paid tools for a stress test. Treat every request using p95-sized tokens as a stress scenario, not a monthly forecast: actual spend depends on how often long requests occur, and input and output p95 may come from different requests.

Starting with a fixed budget instead? Try the chatbot preset with a $100 monthly API budget to estimate how many base requests could fit. Adjust the model and token counts to match your workload. This estimate does not enforce a spending cap.

Pricing data sourced from our catalog. Check data sources for provenance and freshness.

Use the calculator, then inspect candidate models

The calculator gives a monthly estimate for one selected model. The use-case pages and compare tool help you check whether cheaper nearby routes also fit the workload.

Caveats

What these examples do not decide

These examples do not rank model quality, latency, rate limits, cache behavior, discounts, or provider-specific add-on charges. They are starting shapes for monthly API cost estimation. Use the calculator for your own token counts, then confirm final pricing with the provider before production use.