minimax-m3.1POST

M3.1 Flash 跑分：Automation Bench 62.2% 超 Opus 5，DeepSWE 74%

作者列出 M3.1 Flash 跑分：Automation Bench 62.2%（Opus 5 为 50.3%）、DeepSWE 74%、BabyVision 91%、BankerToolBench 76.2 与 Opus 5 持平，10 月 23 日开源权重。

Harish

@Harish_5218 天前收录原推发布于 2026.10.19

Harish @Harish_521原推发布于 2026.10.19

在 X 上打开原推

tbh, i’m really excited about this. MiniMax M3.1 Flash is basically a faster, more refined version of Space Bunny, and the benchmarks look insane: 1. Automation Bench: 62.2% (Opus 5: 50.3%) 2. DeepSWE: 74% 3. BabyVision: 91% 4. BankerToolBench: 76.2 (ties Opus 5) the crazy part? frontier performance at Flash speed and pricing. open weights dropping October 23. bookmark this.

这条是引用着另一条写的：x.com/minimax_ai/status/2112150867868541221

x.com/harish_521/status/2112186753893498934采集于 2026.10.20 20:27