Construction & Engineering

AI models score as low as 2% on construction takeoff test

A construction-specific benchmark found that leading general-purpose AI models performed poorly on drawing interpretation and quantity takeoff tasks.

A benchmark by Togal.AI tested general-purpose AI models on construction takeoff tasks using technical drawings. The tested systems scored between 2% and 34% across 33 tasks involving plan interpretation, element identification, and measurement. Models from OpenAI, Anthropic, Google, xAI, and selected Chinese providers were included. The findings indicate that general-purpose models still struggle with specialized estimating workflows without construction-specific training or tools.

Original source: Construction Briefing

Related RAQI solutions

In-depth analysis on RAQI Insight

Work with RAQI