Moonshot 称 Kimi K3.1 Pro 在 14 项基准中 6 项胜美国前沿模型

Ryan Orhan @rynorhn原推发布于 2026.10.24

在 X 上打开原推

i really wasn’t expecting a chinese lab to close the gap this aggressively right now. moonshot is claiming it beats every us frontier model in its comparison on 6 out of 14 benchmarks. and somehow they’re charging just $2 per million input tokens and $10 per million output tokens. opus 5.5 costs twice that. gpt-6 astra costs FIVE TIMES as much. what the fuck did moonshot cook 😭

这条是引用着另一条写的：x.com/rynorhn/status/2113685242766586208