WireUnWired Research Lab · AI economics
Estimate the cost of your AI workload before it ships.
Compare expected token usage and API spend under your own workload assumptions. Use the calculator below for a planning estimate, then verify the model, rate tier and billing rules on the provider’s official pricing page before budgeting production traffic.
How to read the estimate
Token count is only part of the bill.
Input tokens and generated output may have different prices. Long answers or reasoning-heavy responses can cost more than a prompt-only estimate suggests.
Cached input, batch processing, context-window tiers and regional routing can change the effective price. Verify which tier the calculator assumes.
Images, audio, web search, tool calls, embeddings, storage and other features may be billed separately. Include them when the workload uses them.
A common text-only approximation is:
Estimated cost = (input tokens Ă· 1,000,000 Ă— input rate) + (output tokens Ă· 1,000,000 Ă— output rate)
This formula is only a starting point. Add any applicable cached-token, audio/image, tool, storage or other charge; then account for retries, failed calls, system prompts, conversation history and usage variance. Currency conversion, taxes and credits are not included unless the calculator explicitly models them.
Verify current rates
Use the official price sheet for the final check.
Before choosing a model or presenting a cost estimate to your team, confirm the exact model ID and pricing tier here:
Pricing references checked 29 October 2026. Recheck them whenever a provider changes model availability, discounts or rates. This page provides a planning aid, not a guaranteed quote.
LLM API Cost Estimator
Calculate costs for OpenAI o1, Perplexity Sonar, and Claude with reasoning & caching support.