Unverified50% confidenceOpinionExact time
根据文中量化估算，编程智能体任务中模型推理仅占总时间的 20%-30%，剩余 70%-80% 花在等待外部系统 I/O 响应
1
Sources
50%
Confidence
Long-term
Relevance
8/18/2026
First Seen
Sources
Related Claims
UnverifiedBefore outputting a final answer, a reasoning model's internal chain of thought may consume 3,000 to 20,000 tokens, far exceeding the approximately 500 tokens of an ordinary model.73% similarUnverifiedOn cognitive tasks, humans can achieve approximately 80%, while SOTA multimodal models achieve only 30% to 50%, sometimes approaching random guessing72% similarUnverifiedWhen solving complex programming problems, the number of internal reasoning tokens in the o1 model is often 10 to 50 times the final output.72% similarUnverified大模型的推理成本与参数量大致成正比，而实际业务中约80%的任务可以由中小规模模型胜任72% similarUnverified推理模型通过思维链或内部搜索进行多步推演，token消耗量可能是普通模型的5到20倍71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/766340API
curl https://kongchang.com/api/v1/knowledge/claims/766340MCP
get_claim(id=766340)