Savings
Prompt caching savings calculator
What prompt caching saves on a repeated prefix, including the cache write cost where the vendor charges one.
Prices read from the vendor pages on 2026-10-20. Arithmetic on list prices, not advice or a quote.
Saving and exports need Norwell Tools Pro, $9 a month. See what is in it. Everything above is free.
How it works
Caching pays when the same long prefix is sent often within the cache lifetime. Anthropic charges a write premium (25 percent over input for the 5-minute cache) and bills reads at 5 to 10 percent of input; OpenAI bills cached input at a discount with no write charge. Misses pay the write price here.
The formula
with cache = hits * (prefix * read_price) + misses * (prefix * write_price) + all * (variable * input_price + output * output_price); write_price = input_price where the vendor charges no write premiumWorked example
With the inputs the page opens on:
- model: claude-sonnet-5-5
- cached prefix tokens: 20000 tokens
- variable tokens: 500 tokens
- output tokens: 400 tokens
- requests per month: 50000
- cache hit rate pct: 90%
gives
- monthly without cache: $2,250
- monthly with cache: $590
- monthly savings: $1,660
- savings pct: 73.8 %
Sources
- https://developers.openai.com/api/docs/pricing (read 2026-10-20)
- https://platform.claude.com/docs/en/about-claude/pricing (read 2026-10-20)
- https://ai.google.dev/gemini-api/docs/pricing (read 2026-10-20)
Prices read from each vendor's own pricing page on 2026-10-20; the page may have changed since. This is arithmetic on published list prices, not advice or a quote. Check the vendor's page before committing a budget.