deepseek-v4.1POST
上手反馈：V4.1 Flash 聪明但想太多、光想不做
用户评价 DeepSeek V4.1-Flash 1017「进五步退两步」，模型聪明却过度思考，倾向于反复推理而不动手执行。
unbug
@unbug11 天前收录原推发布于 2026.10.18
unbug @unbug原推发布于 2026.10.18
DeepSeek V4.1-Flash 1017 feels like five steps forward, two steps back. Smart, but it overthinks. It thinks instead of doing.
这条是引用着另一条写的：x.com/unbug/status/2110025934338408704
x.com/unbug/status/2111517136820867434采集于 2026.10.18 16:29