Methodology page draft

How the estimate works

The core calculator uses a simple token-cost model:

((requests × input_tokens × input_price) + (requests × output_tokens × output_price)) / 1,000,000 × buffer

Why we use a simple model first

This site is meant for early planning. A simple, inspectable model is often more useful than a fake-precise dashboard that hides assumptions.

What this model includes

• request volume
• average input size
• average output size
• editable token pricing
• optional retry or safety buffer

What this model does not fully include

• prompt caching differences
• retrieval or embedding costs
• tool execution costs
• image, audio, or video generation charges
• moderation and evaluation traffic
• pricing tiers, credits, or enterprise discounts

Best practice

Use this estimate as a starting point, then adjust based on logs, pilot traffic, and provider billing details.

Recommended trust module

Add a "How we estimate" box beside results with links to Sources and Disclaimer.