AREX agents

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

DeepLearning.AI(@DeepLearningAI) · 6.5 分

GLM-5.3模型在漏洞利用测试中表现接近Claude Mythos，性能差距仅2%。文章还提及其他AI模型进展与技术动态。