llmcosts

Simple web app to estimate LLM spend from tokens per call and request volume. State is shareable via URL params.

Demo

llmcosts.vercel.app

Why

LLM features look cheap at prototype scale and expensive at production scale.
llmcosts gives quick order-of-magnitude clarity.

How it works

Cost per call = (input_tokens / 1e6 * input_price_per_1M) + (output_tokens / 1e6 * output_price_per_1M)
Billable calls/day = calls/day * (1 - cache_hit_rate)

Pricing

Edit src/lib/pricing.ts. Values are placeholders.

Run locally

npm install
npm run dev