Skip to content
AI HOT (Curated Pool)

Claude Mythos Evaluation Shows 16-Hour Risk Horizon

Claude Mythos评估显示16小时风险时距

METR evaluated an early Claude Mythos Preview build during a limited March 2026 window and estimated its 50% time horizon at at least 16 hours, with a 95% confidence interval of 8.5 to 55 hours.

Why it matters: METR's risk assessment of an early Claude Mythos preview measured a 16-hour lower bound on tasks the model can finish alone.

Read the original ↗Export Markdown