Unverified50% confidenceFactExact time
The prompt caching mechanism allows already-processed input prefixes to be cached server-side, and if the next request starts with the same prefix, it is read at a price far below normal input; common cache TTLs are 5 minutes and 1 hour, and writing to the cache costs extra.
1
Sources
50%
Confidence
Long-term
Relevance
10/16/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedPrompt caching means the identical prefix shared by multiple requests is fully billed and processed only on the first request, and subsequent requests hitting the cache can lower the input token price and reduce latency; but caching cannot shorten the total context length, and long histories still occupy the context window and raise latency.73% similarUnverifiedCached input refers to pre-storing the same context on the server side so that subsequent requests can reuse the cache without repeated computation, and is billed at a lower price.70% similarUnverifiedThe new version adds hints about possible reasons for cache misses in the prompt_cache field of the /cost command and the status bar, such as changes to tool definitions or system prompts, or idle time exceeding the TTL69% similarUnverifiedPrompt caching usually has an expiration period, requires rewriting after content changes, and the cache discount applies only to the input side, while output tokens are still billed at the standard rate.68% similarVerifiedThe cache read mechanism retains the KV cache for recurring long prompt prefixes, charging at an extremely low rate upon a cache hit; the leaked cache read price for Opus 5.5 is approximately 1/20 of the input price.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1047798API
curl https://kongchang.com/api/v1/knowledge/claims/1047798MCP
get_claim(id=1047798)