METR evaluates autonomy on long research tasks before model release
METR evaluated Claude Opus 5.5’s ability to accelerate AI research on difficult, long-horizon tasks. Its preliminary assessment judged the model unlikely to fully automate AI research.
PRACTICAL IMPLICATIONS
What does this mean?
Success on individual tasks differs from running an entire research process independently. Evaluations help track the gap between these capabilities.
LIMITS OF THE EVIDENCE
What does this not establish?
The assessment was preliminary and testing time was limited. It did not evaluate every safety property. The work was conducted under an unpaid agreement with Anthropic, which could comment on the summary.
Significance: significant
The assessment helps distinguish task-level capability from full research automation.
An editorial judgement, not a numerical threat forecast.
ORIGINAL PUBLICATION
Summary of METR's predeployment evaluation of Claude Opus 5.5
METR · published 22 Sep 2026
Read the original source ↗Source checked 13 Nov 2026. Summary updated 13 Nov 2026. The original publication date is not necessarily the date of the event. This summary was written with AI assistance using the original source.
← Back to the archive