Prompt caching savings calculator
Compare what a repeated system prompt, tool list or document costs with and without prompt caching.
Model
Claude Sonnet 5.5: ₹195.07 per 1M input tokens, ₹19.51 per 1M cached input tokens.
Your prompt
System prompt, tool definitions, documents
The part that changes each time
Traffic
The first request, and one after each gap long enough for the cache to expire
1.25 for Claude; 1 where caching is automatic and writes cost the same as input
Per day on Claude Sonnet 5.5
Rupee prices are read from the live catalog, last updated 2026-11-22. They follow the USD/INR rate, so a real bill can differ slightly.
Get an API keyHow the saving is worked out
Without caching, the repeated part is charged at the input price on every request. With caching, it is charged at the write price on the requests that create the cache and at the cached-input price on the rest. New input and output cost the same either way.
When it does not help
A cache entry expires after a few minutes without a hit, so low or bursty traffic means more writes. Short prompts may be below the provider's minimum cacheable size. The prompt caching docs cover expiry, minimum size and how to turn it on.
Check it on real traffic
Your usage log shows cached input tokens per request when the provider reports them. For a whole workload rather than one prompt, use the LLM cost calculator, which takes a cached share of the prompt.