GLM-5.3模型接近Claude Mythos安全表现
This week in The Batch, Andrew's Letter covers how open weights model GLM-5.3 nearly matched Claude ...
精选理由
GLM-5.3安全性能接近Claude Mythos，开源模型又进步了，还有小米和DeepSeek的新消息。
开源模型GLM-5.3在利用漏洞方面的表现达到Claude Mythos的86%，12%比14%。小米发布了新的开源权重模型领导者。Gemini 3.8 Live版本已发布。DeepSeek-V4.1-Flash模型更新。AREX智能体中心上线。
原文 · DeepLearning.AI
This week in The Batch, Andrew's Letter covers how open weights model GLM-5.3 nearly matched Claude ...
This week in The Batch, Andrew's Letter covers how open weights model GLM-5.3 nearly matched Claude Mythos at exploiting vulnerabilities, 12% vs. 14%. Also inside: 🧪 Xiaomi's new open weights leader 🎙️ Gemini 3.8 Live 💾 DeepSeek-V4.1-Flash 🔁 AREX agents hubs.la/Q04z20mv0 Hm #DeepLearningAI i #AIEngineering e #OpenWeights ights 💬 5 🔄 0 ❤️ 12 👀 1433 📊 6 ⚡
