Tom's Hardware · AI· Sayem Ahmed ·· 2 天前AI 评分62
Anthropic声称国产模型GLM-5.3具备Mythos级黑客能力
Anthropic claims popular Chinese AI model has Mythos-class hacking abilities
AI 导读
Anthropic发布报告称Zhipu AI的GLM-5.3在Exploitbench中实现端到端漏洞利用50次/410次，接近其未发布的Mythos模型；在内部
来源：Tom's Hardware · AI · tomshardware.com