Kimi K3.1 Pro Ranks Fourth, Then Apologizes for Quota Issues
Moonshot AI released Kimi K3.1 Pro as an upgrade to its July K3 model, with free-to-download weights planned by Nov. 4. Artificial Analysis reportedly scored it 55 on its Intelligence Index, placing it fourth overall and first among open models, while some users reported slow performance and occasional wrong paths. Within 48 hours, Kimi apologized for rate limits and unnecessary quota consumption among paying users, offering compensation equal to one week of usage.
Kimi K3.1 Pro combines a top-four independent benchmark placement with reported problems in paid-user quota consumption, highlighting the gap between model rankings and the practical experience of its long-running agent features.
日前，Kimi发布K3.1 Pro。发布后48小时，Kimi为一周额度发了致歉信。
一边是8个月，估值从43亿美元冲到500亿美元的狂欢。一边是付费用户的账单。两个针对不同对象的预期管理，在同一个时间窗口撞在了一起。
Kimi需要向市场回答或者证明一个问题：用户愿意为全球第四的跑分，持续买单。毕竟所有的估值，最终由用户的账单决定。
两个预期管理
“K3.1就是我心中的国产第一。但仍有两个点待提升，速度比较慢，额度依然很少。”
“K3让人有种惊艳感，K3.1 Pro步骤很密但还是有概率把路走岔了，走错一步就真的浪费token和时间。”