minimax-m3.1POST
BankerToolBench:M3.1 Flash 超过 DeepSeek、Kimi、GLM,追平 Opus 5
作者称在自研 BankerToolBench 上 MiniMax M3.1 Flash 胜过 DeepSeek、Kimi 和 GLM,与 Opus 5 持平,认为高价值专业场景未必再需要昂贵闭源模型。
Jonas Mueller
@jomulr8 天前收录原推发布于 2026.10.21
Jonas Mueller @jomulr原推发布于 2026.10.21
On BankerToolBench, MiniMax M3.1 Flash outperforms DeepSeek, Kimi, and GLM, matching Opus 5. High-value professional AI may not need expensive closed-source models for much longer. Our BankerToolBench paper was also accepted at #NeurIPS2026. See you in Sydney!
这条是引用着另一条写的:x.com/minimax_ai/status/2112150867868541221
x.com/jomulr/status/2112587089334403313采集于 2026.10.21 12:04