SemiAnalysis launches benchmark to evaluate AI agents on complex tasks
Models & ResearchSuperintelligence · 2h ago

SemiAnalysis launches benchmark to evaluate AI agents on complex tasks

SemiAnalysis created the GDPx benchmark to measure model performance on difficult technical challenges, including chip engineering and quantitative trading. Claude Opus 5.5 achieved the highest score at 12 percent, with researchers noting that built in safety filters reduce top scores across cybersecurity tests.

SemiAnalysis