跳到正文
DeepLearning.AI· DeepLearningAI·· 16 天前AI 评分28

The Batch周报:GLM-5.3与Claude Mythos对比

This week in The Batch, Andrew's Letter covers how open weights model GLM-5.3 nearly matched Claude Mythos at exploiting vulnerabilities,...

AI 导读

本周《The Batch》中,Andrew 的信涵盖开源权重模型 GLM-5.3 在利用漏洞方面几乎追平 Claude Mythos(12% vs. 14%)。此外还包括:小米新的开源权重领导者、Gemini 3.8 Live、DeepSeek-V4.1-Flash、AREX agents

来源:DeepLearning.AI · x.com