Anthropic
Anthropic Confirms: Zhipu GLM-5.3 Closes the Exploit Gap with Claude
GLM-5.3 scored 4% on binary exploits vs Claude Mythos Preview's 6%; prior models 0%. Red Team test shows advanced AI attack is now US-China parity.
Back to home
Claude Mythos
3 articles tagged with this topic
Claude Mythos
Claude Mythos Locked Down, Alibaba's HappyHorse Tops Video Charts
Anthropic withholds its strongest model from public access; Alibaba's anonymous video model tops benchmarks before reveal.
Apr 124 min readopcnew Editorialjuejin.cn
Anthropic
Anthropic Restricts Claude Mythos to Security Researchers Only
Claude Mythos can chain 4-vulnerability browser exploits autonomously; Anthropic limits access via Project Glasswing.