Matrak releases new benchmarks for AI takeoff and estimation agents

Matrak has released updated benchmarks and a system card addendum for its AI Takeoff and Estimation system. The report compares the performance of its Matrak Connect Agent against five frontier models, including ChatGPT 5 and Fable 5, across six live projects using audited ground truth and agentic conditions.

The AI Takeoff Agent is designed to automate the process of counting and measuring products from 2D drawings to create tender-ready quotes. This task, known as a ‘takeoff’, is currently a high-stakes, labor-intensive activity. In the US, UK, and Australia alone, the economic cost of this process involves over 500,000 full-time roles, with the number approaching one million when including part-time and outsourced estimators.