The Decoder· Maximilian Schreiner·· 1 d ago
Anthropic says Zhipu's open-weight GLM-5.3 nearly matches Claude Mythos Preview at building exploits
Anthropic says Zhipu's open-weight GLM-5.3 nearly matches Claude Mythos Preview at building exploits
AI summary
Anthropic published an evaluation saying Zhipu's open-weight GLM-5.3 model approaches its Claude Mythos Preview in exploit development.
Selection record
Threshold 76Media and individualsFirst 78Second 78
AdmittedSum of both 156 ≥ twice the threshold 152
- Source tier
- Media and individuals; this tier's threshold is 76
- Pre-filter
- passed：正文详述GLM-5.3模型漏洞利用能力评测
- Why it was chosen
- Anthropic compared GLM-5.3 with Claude Mythos Preview using two benchmarks and one practical vulnerability discovery test, and gave specific costs for removing open-weight guardrails.
- Same event
A model scores each item twice, independently, against one written standard, out of 100. An item is admitted only when the two scores add up to twice the threshold. The threshold is set per source tier.
Source: The Decoder · the-decoder.com