Anthropic Frontier Red Team：GLM-5.3 与 Claude Mythos Preview 在 Binary Exploitation 基准上的表现
Quoting Anthropic Frontier Red Team
Anthropic Frontier Red Team 在内部 Binary Exploitation 基准的 100 个随机任务上评测多个模型，GLM-5.3 在 4% 的试验中实现完整的控制流劫持，Claude Mythos Preview 为 6%。报告称这一结果意味着一个有意义的能力门槛已被跨过，此前的 Claude Opus 4.6 和 GLM-5.2 在这些任务中均未成功。
29th September 2026
We evaluate several models on 100 tasks from the [internal Binary Exploitation benchmark] (selected at random), and find that GLM-5.3 develops full control flow hijacks in 4% of the trials; Claude Mythos Preview did so in 6%. Although GLM-5.3 performs below Claude Mythos Preview here, a meaningful threshold has clearly been crossed: earlier models, like Claude Opus 4.6 and GLM-5.2, do not succeed in any of them.
— Anthropic Frontier Red Team, GLM-5.3 and the spread of advanced cyber capabilities
来源：Simon Willison · simonwillison.net