Expired60% confidenceFactTime unknown
According to Anthropic's official empirical data, multi-agent architecture may consume 15 times the number of tokens compared to ordinary single-agent conversations.
2
Sources
60%
Confidence
Medium-term (~90 days)
Relevance
7/2/2026
First Seen
Valid until: 9/30/2026(expired)
Sources
Related Claims
Unverified智能体任务的Token消耗量通常是普通对话的数十倍71% similarUnverified深度研究单次任务消耗的token数量可达普通对话的数十倍，成本可能相当于数十次乃至上百次普通问答的资源消耗66% similarUnverified多Agent协作意味着多次模型调用，Token消耗成倍增长63% similarUnverifiedCascade模式下携带完整项目上下文的长对话，单次请求的输入token数可能达到数万甚至十万级别61% similarUnverified对话历史积累构成一种转换成本，用户积累的上下文越丰富迁移代价越高，平台黏性越强58% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/60541API
curl https://kongchang.com/api/v1/knowledge/claims/60541MCP
get_claim(id=60541)