Summary of METR's predeployment evaluation of Claude Opus 5.5

researchMETR ResearchSep 22, 2026

Note on independence: This evaluation was conducted under an unpaid agreement for AI R&D assessment. 1 We drafted the initial summary, and then Anthropic had the opportunity to review and edit the text. We signed off on this final text from the Claude Opus...

Impact: MediumConfidence: MediumNovelty: Incremental
researchsource-feed-92851532-c6c4-4203-933d-ecc01d4e1171agentsmultimodalsafety

See the whole story

Read the original on METR Research

Sens.ai aggregates and assesses published reporting. The assessment above is machine generated; the original source is authoritative.