Misusing Anthropic Caching Can Actually Cost More
KitchenAmoeba4438 · reddit · 2026-07-18
After investigating an unusually high Anthropic API bill, the author discovered a commonly misunderstood detail: Anthropic has its own caching mechanism, and messages must be an "exact match" with previous content to hit the cache and be billed at cache token rates.
They argue that many context-compressing plugins might actually be more expensive in Anthropic's ecosystem:
- Whenever a plugin needs to "rehydrate" the context, it may consume extra tokens first
- If it alters the cache structure expected by the Anthropic provider, it could lose cache token pricing and revert to more expensive standard token rates
- Different providers implement Anthropic caching differently, adding further complexity
The conclusion is that saving tokens on the Anthropic API isn't straightforward; some tools that appear to save context might actually inflate your bill.
More from coding & agent
- Reticle builds a verification layer that lets AI coding agents check their own generated code — Scobleizer · 2026-10-27
- Dev training a guard model to block prompt injection, tool and memory poisoning — gnukeith · 2026-10-27
- Dev trains a ~4GB-VRAM sentinel model to guard against prompt injection and poisoned memory — gnukeith · 2026-10-27
- Superset rebrands as Paradev, raises $11.5M seed led by USV — massimosgrelli · 2026-10-27
- Ex-Big Tech managers may thrive in the AI era: manage agents like you managed teams — lxfater · 2026-10-27
- Claude debugs airline and hotel sites; Fasano's date picker is all NaN and invalid dates — lilyraynyc · 2026-10-27