AnalysisPolicyAugust 25, 2026

Reddit users question METR's benchmark inactivity

A Reddit post criticizes METR for abandoning its benchmark after Opus 4.6, despite receiving millions in investments. The user argues METR could still benchmark current models for 80/90/99% accuracy with its existing task set.

1 source

Daily brief

Get tomorrow's AI brief in your inbox

More stories today

Open the live feed