Expired50% confidenceFactExact time
Across 5,546 successful chat calls from 16 customer accounts, the median request carried 1,280 input tokens and returned 108 output tokens.
1
Sources
50%
Confidence
Short-term (~14 days)
Relevance
10/19/2026
First Seen
This claim has expired. It is kept for the record and no longer describes the current state.
Valid until: 11/2/2026(expired)
Sources
Related Entities
Related Claims
UnverifiedIn a typical scenario of 20 rounds of interaction, if the system prompt occupies 15,000 tokens and each round produces an average of 2,000 tokens, by the 20th round the input token count for a single request balloons to 55,000 tokens.70% similarUnverifiedThe project spanned 55 sessions, with 40,473,602 input tokens, 1,587,482 output tokens, and 561,271,443 cache read tokens.70% similarUnverified在一个典型项目查询中发送了45,000 tokens的上下文，但真正有用的只有约5,000 tokens68% similarUnverified一次典型的RAG调用可能检索5-10个文档块，每块500-1000个Token，输入Token总量可达5000-10000个，而生成回答通常只有200-500个Token65% similarUnverifiedAccumulating from May 15, 173 snapshots generated 1,889 queries, of which 605 received no answer at all, accounting for 32% of all queries.63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1067109API
curl https://kongchang.com/api/v1/knowledge/claims/1067109MCP
get_claim(id=1067109)