07:00
2026-09-22
metr.org
ai-safety
Summary of METR's predeployment evaluation of Claude Opus 5.5
METR's preliminary predeployment evaluation of Claude Opus 5.5 concluded that the model would accelerate AI R&D slightly more than Fable 5.1 but is unlikely to fully automate AI R&D, based on capabili…