Unverified50% confidenceFactExact time
OpenAI和Anthropic均已推出Prompt Cache功能，可复用KV Cache，对调用方体现为token费用折扣（通常为50%–90%）
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/10/2026
First Seen
Valid until: 10/8/2026
Sources
GPT-5.6曝光：Agent编程token效率提升54%意味着什么
hackernewshackernews7/9/2026
Related Claims
UnverifiedOpenAI于2024年推出Prompt Caching功能，当提示词前缀完全相同且超过1024 token阈值时复用KV Cache，缓存部分通常按原价25%甚至更低计费76% similarUnverifiedAnthropic和OpenAI均已推出Prompt Caching相关折扣政策74% similarVerifiedOpenAI Prompt Caching 可将重复前缀的缓存成本降至原价的10%-25%72% similarUnverifiedAnthropic的Claude API和OpenAI的API均已原生支持提示缓存功能，缓存命中时的输入Token价格通常仅为正常价格的10%-25%72% similarUnverifiedPrompt Caching通过复用前缀KV Cache，以Anthropic定价为参考可将延迟降低约85%、成本降低约90%72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/459758API
curl https://kongchang.com/api/v1/knowledge/claims/459758MCP
get_claim(id=459758)