Expired50% confidenceBenchmarkExact time
After state trimming, the input tokens were reduced from over 20,000, and the model's judgments on menus and blind selections became accurate.
1
Sources
50%
Confidence
Short-term (~14 days)
Relevance
9/19/2026
First Seen
This claim has expired. It is kept for the record and no longer describes the current state.
Valid until: 10/3/2026(expired)
Sources
Related Entities
Related Claims
Unverified在设计不良的多智能体系统中，协调 Token（Agent 间传递的指令、确认、状态摘要）可占总消耗的40%-60%65% similarUnverified同一个精心设计的提示词在不同平台间直接复用时，输出质量可能出现20-40%的效果落差64% similarUnverifiedThe model's local throughput is approximately 9 to 22 tokens per second, consumes more tokens than the previous generation, and is slower when thinking mode is enabled.64% similarUnverifiedThe officially claimed 30% reduction in Thinking Token consumption only held true for two of the five problems; for the remaining three, K2.7-Code actually consumed more tokens.63% similarUnverifiedIncluding specific evaluation dimensions in prompts (such as format, word count, key points to cover, and items to exclude) can improve model output quality by an average of 30%–50% compared to vague prompts.62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/933999API
curl https://kongchang.com/api/v1/knowledge/claims/933999MCP
get_claim(id=933999)