Unverified50% confidenceFactExact time
After enabling Extended Thinking in the Claude series, token consumption per request can be several times that of normal mode.
1
Sources
50%
Confidence
Long-term
Relevance
10/1/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedModern Claude models already execute behaviors such as 'check again' and 'be as thorough as possible' by default, and repeatedly requesting these instructions doubles costs while producing worse answers.69% similarUnverified随着对话轮次增加，每轮需携带完整历史上下文，输入token数量持续膨胀，导致推理时间呈现非线性增长61% similarUnverified模型有时会陷入过度思考，在特定任务中产生超过52倍的额外token60% similarUnverifiedAs large model reasoning capabilities strengthen, the tokens consumed per single call rise significantly, creating a paradox of 'stronger model' versus 'faster limit reached' under fixed usage limits.60% similarUnverifiedClaude Code的向量索引替代方案在大上下文窗口（200K token）成熟后才变得可行，当模型能一次性阅读数十个文件时按需查阅比预计算索引更精确实时60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/965295API
curl https://kongchang.com/api/v1/knowledge/claims/965295MCP
get_claim(id=965295)