Skip to content
Trending storyDeveloping

Google's PLE benchmark: best AI agent scores 50.6% on physical engineering tasks

1 article1 sourcesince Oct 28Last article 1h ago ·

Overview

AISummary of 1 article

Google researchers introduce PLE, a benchmark of 85 long-horizon physical engineering tasks in electrical, civil, structural and mechanical fields, written or checked by practising engineers.

Claude Opus 5.5 in Claude Code scores highest of 14 models, with a mean of 50.6%, using 27M tokens and 63 minutes per task.

On design tasks, the best agent's outputs met 91% of constraints, but only 41.3% met every hard constraint, according to the paper as reported by Rohan Paul on X.

Written by AI from the articles below · updated Oct 28, 6:18 PM ET

Check the sources:

Latest developmentOct 28, 6:11 PM ET
Google's PLE benchmark finds the best AI agent scores 50.6% on physical engineering tasks

Article timeline

The articles in this story. Times are ET.

Oct 28
  1. Rohan PaulX
    Google's PLE benchmark finds the best AI agent scores 50.6% on physical engineering tasks

    AIGoogle researchers introduce PLE, a benchmark of 85 long-horizon physical engineering tasks written or checked by practising engineers. The best configuration, Claude Opus 5.5 in Claude Code, scores a mean of 50.6% across 14 models. On design tasks, outputs met 91% of constraints but only 41.3% met every hard constraint.

    Image from @rohanpaul_ai's post

Heat Trend

Not enough continuous observations to show a trend yet.