Unverified50% confidenceFactExact time
When contextual retrieval uses prompt caching, the cost is about $1.02 per million document tokens, one-time.
1
Sources
50%
Confidence
Short-term (~14 days)
Relevance
10/18/2026
First Seen
Valid until: 11/1/2026
Sources
Related Entities
Related Claims
UnverifiedBy estimate, when an 8000-token document is split into 10 chunks (each with 100 tokens of context) and caching is used, each document costs about $0.003, i.e. about $0.38 per million document tokens; without caching it is about $1.17 per million tokens.78% similarUnverifiedIn the warm cache case, a /compact request on a 136.4k token context costs about $0.056–$0.059, which is 1.8–1.9 times that of a normal turn (about $0.031).77% similarUnverifiedA typical RAG query (2000-token prompt, 2500-token retrieved context, 400-token output) costs about $0.013 on a mid-tier model.74% similarUnverifiedAt $1 per million tokens, the inference cost for 500 billion tokens is approximately $500,000.74% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1063910API
curl https://kongchang.com/api/v1/knowledge/claims/1063910MCP
get_claim(id=1063910)