WHAT NEW IN X10.28 WED

X 上的 AI 合集档案 · 22 个合集 · 58,924 条

minimax-m3.1IMAGE

实测:浏览器/电脑使用基准中 M3.1 Flash 与 Sonnet 同分却未被推荐

作者在 Runable 的 computer/browser use 基准里发现 MiniMax M3.1 Flash 与 Anthropic 模型得分完全相同,Sonnet 还失败了一个任务,但工具仍判定 Anthropic 更优,因此质疑评测偏向。

Umesh Kumar
@itsumeshk2 天前收录原推发布于 2026.10.25
实测:浏览器/电脑使用基准中 M3.1 Flash 与 Sonnet 同分却未被推荐
Umesh Kumar @itsumeshk原推发布于 2026.10.25
在 X 上打开原推

When you do any kind of testing with Anthropic models, it almost always find out ways to tell that anthropic is the best, I am not sure i can trust Anthropic models now. In a computer/browser use bench we use for Runable MiniMax M3.1 Flash scored exactly the same but still its not recommending it. Sonnet actually failed at one task but still somehow its better.

x.com/itsumeshk/status/2114366724002967561采集于 2026.10.26 12:06