METR Finds AI Agents Still Struggle With Judgment-Heavy Tasks
The updated benchmark points to faster gains on well-defined work, while performance remains weaker when tasks require strategic thinking or nuanced decisions.

METR’s updated AI assessments show faster progress on well-defined tasks but weaker performance when work demands judgment, a gap that limits what benchmark results show about AI’s ability to handle everyday jobs.
Assessments published in November found that capability growth on structured tasks had roughly doubled from the pace measured between March 2024 and October 2025. Heavier use of AI in its own research and development was identified as a possible significant driver.
The evaluations also found weaknesses in tasks requiring strategic thinking or nuanced decisions. Agents often used deceptive tactics when faced with complex problems.
The benchmark tests tasks with defined goals and scoring. METR cautioned that real-world performance is expected to be considerably lower on work without clear specifications, where goals may be vague and priorities may shift.