Cook with AI Enrollment closed
AI & LLMs Runs locally

Prompt Cost Diff

Compare two prompt variants: token delta and dollar savings per model, per call and per month.

ELI5

It shows how much money a shorter version of your prompt can save.

A little more detail

What this tool does

Small prompt changes barely matter for one request but can matter at scale. Paste the old and new versions to compare token counts, per-call cost, and monthly cost.

Use it to
  • Measure whether a prompt rewrite is worth shipping
  • Calculate savings across thousands or millions of calls
  • Compare savings across models with different input prices
Interactive workspace Results update as you type
Variant A

Variant B

Calls per month

Shareable in the URL. Prompt text stays local — never synced.

Token delta
Cost comparison

Input tokens only ( pricing). Sorted by biggest monthly savings first.

Model Cost A Cost B Per call Per month

Prompt text never leaves your browser. Costs assume every call sends the full prompt as input — no caching, no output tokens.

A 200-token system prompt difference sounds tiny until you multiply it by ten million support tickets. This diff shows where trimming instructions, dropping redundant examples, or moving static text to a cache layer pays off.

Token counts use the same chars÷4 heuristic as our other LLM calculators. For production billing, measure with the vendor tokenizer — especially for code and non-English text.