Prospective Evaluation of Large Language Models on the 2026 FIFA World Cup

Large language models have achieved superhuman performance on standardized examinations based on accumulated knowledge, yet their ability to forecast real-world future events remains largely unexplored. We conducted a prospective evaluation of eight leading language models, four human experts, and betting-market odds on the 2026 FIFA World Cup, eliciting predictions for quarter-finalists, semi-finalists, champion, Golden Ball, and Golden Boot after the Round of 16 was set but before knockout matches began.