investigation
Turning off thinking entirely (None level) is sometimes more expensive than Low, because the agent compensates with a four-times-longer action chain that inflates the context window and input token bill.
A investigation by החוקרים. Statements are as reported by the cited publishers — not verified by DecisionGateway.
ואל תנסו להתחכם ולכבות את החשיבה לגמרי (רמת None): זה לפעמים יקר יותר מ-Low, כי הסוכן מפצה על חוסר התכנון בשרשרת פעולות ארוכה פי ארבעה שמנפחת את חלון ההקשר ואת החשבון על טוקני הקלט.
Discussion
Related claims
Raising reasoning effort improves average benchmark scores but produces painful performance drops at the individual task level; tasks solved easily and cheaply at Low fail at High or xHigh.החוקרים · investigationBetween 58% and 98% of the true cost of running an agent at high effort is not the AI bill but the opportunity cost of developers' time waiting for the model.החוקרים · investigationOn easy tasks solved at every effort level, costs can be up to 42x higher in API costs because the model treats effort as a fixed token multiplier.החוקרים · investigation'The end of diffusion' is hype; the better question is whether it is worth paying for more expensive training to get built-in verification and flexible compute at inference time — a real and interesting trade-off.מייק · statementThe researchers propose naive laddering: always start at low effort and escalate to high only on failure, since models are bad at self-routing in advance; this cuts costs by tens of percent and improves performance.החוקרים · commitmentThe message cost calculator includes a worked example showing how a good bot cuts the bill by 30% and closes the conversation in two messages.post author · statement