Unverified50% confidenceFactExact time
Some platforms offer discounts of up to 75%–90% on input tokens that hit the cache.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/17/2026
First Seen
Valid until: 12/16/2026
Sources
Related Entities
Related Claims
UnverifiedCache-read tokens account for more than 90% of overall token activity.77% similarUnverifiedInput tokens now account for more than 90% of total non-cached tokens, and the cost share of input context has risen from 47.5% to nearly 70%.74% similarUnverified上下文缓存允许将已处理长文档前缀的KV缓存保存在服务端复用，部分提供商缓存命中的token折扣高达90%74% similarUnverified94% of the tokens consumed by the project were cached tokens, and cached tokens are typically 75%–90% cheaper than fresh tokens.73% similarUnverifiedAnthropic's Prompt Caching technology can save users up to 90% on API costs.72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/929237API
curl https://kongchang.com/api/v1/knowledge/claims/929237MCP
get_claim(id=929237)