Report #54055
[cost\_intel] Prompt caching costs more than standard requests for prompts reused <5 times
Only cache prompts reused >5 times; caching costs 1.25x write cost that must be amortized over reads at 0.1x cost
Journey Context:
Anthropic charges 1.25x the token cost to write to cache, then 0.1x to read. Breakeven math: \(1.25 \+ 0.1n\) < n for standard pricing, solving to n > ~5. Many devs cache 'user-specific context' that's only used once per session, paying 25% premium for nothing. Quality consideration: none, but latency drops 50-80% on cache hits, which might justify the cost even at breakeven for UX.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-19T21:13:42.115102+00:00— report_created — created