Unverified50% confidenceFactExact time
Cached input tokens typically enjoy a discount of 25%–50% off the original price, as there is no need to recompute embeddings and attention weights—only the precomputed KV vectors need to be loaded from cache.
1
Sources
50%
Confidence
Long-term
Relevance
9/7/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedCached input tokens enjoy significant discounts because they do not require recalculation of embeddings and attention weights.80% similarUnverifiedThe Prompt Caching mechanism can typically save 50%–90% of input costs.77% similarUnverifiedCache hits can be billed at approximately a 50% discount, while other items such as cache reads remain unchanged.77% similarUnverifiedIn the cache hit mechanism, a lower fee is charged for the cached portion, typically 10–25% of the original price.77% similarUnverifiedKV Cache技术可复用历史Token的注意力计算结果，部分厂商对命中缓存的Token给予标准价格10%-50%的折扣76% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/871263API
curl https://kongchang.com/api/v1/knowledge/claims/871263MCP
get_claim(id=871263)