标准模型就是牺牲准确率提高速度，跟让一个人不思考零秒快答一样，跟蠢蛋无关。

GPT-5 Thinking is a mode of the base GPT-5 architecture that emphasizes depth and accuracy over speed for complex queries, using more computational resources and time to provide a thorough, well-reasoned answer, much like the model switching to a slower, more detailed mode for difficult tasks. The standard GPT-5, in contrast, prioritizes speed, providing quick, off-the-cuff answers for most common requests.  

所有跟帖： 

• more detailed mode不是提高了正确性, 而是提供更多细节, 现在我的感觉就是说更多废话 -testmobile- ♀ (0 bytes) () 09/08/2025 postreply 11:05:05