CladBench – an open benchmark for AI on UK building regulations
Hacker News
Read full postCladBench is an open benchmark evaluating large language models on UK and EU building regulations with 536 questions across 12 categories. Seven models were tested, with Claude Opus 4.7 scoring highest and Gemini 2.5 Pro and GPT-5 performing comparably. The benchmark highlights varying model performance depending on task category rather than overall score.



