Unverified50% confidenceFactExact time
In long-running agentic conversation scenarios, more than 90% of input tokens are billed at the cache price.
1
Sources
50%
Confidence
Long-term
Relevance
9/23/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe Prompt Caching mechanism can typically save 50%–90% of input costs.72% 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%.70% similarUnverifiedInput tokens and output tokens are typically priced differently, and in multi-turn conversations, repeatedly passing in historical messages causes costs to accumulate linearly with conversation depth69% similarUnverifiedIn intelligent programming scenarios, input tokens often account for more than 80% of total consumption.69% similarUnverifiedAutonomous coding agents consume far more tokens than ordinary conversational scenarios, so reductions in API input/output unit pricing will compress overall operational costs in a nearly linear manner.68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/942279API
curl https://kongchang.com/api/v1/knowledge/claims/942279MCP
get_claim(id=942279)