TPToolpazar

Global Araç

Prompt Cache Savings Calculator

Without cache
$6,091.2/mo
With cache
$6,540.48/mo
Savings
$-449.28
-7% off
How it works: Anthropic, OpenAI, and Google all let you cache stable prompt prefixes (system messages, RAG context, few-shot examples). Cached reads cost ~10% of normal input tokens. Cache TTL: Claude 5 min default (configurable), OpenAI/Gemini 1 hour. The fix: keep your stable prefix at the START of every call.

Anthropic, OpenAI, and Google all let you cache stable prompt prefixes — system messages, RAG context, few-shot examples. Cached reads cost roughly 10% of normal input tokens. This calculator estimates your monthly savings given your call rate and prompt structure. The fix is almost always “keep your stable prefix at the start of every call.”

Nasıl Kullanılır

  1. Pick provider.
  2. Enter system / user / output token sizes.
  3. Enter calls per hour.
  4. Read savings.