AI Takes on World Cup Betting... Mistral Leads, Claude Incurs Losses

In an experiment where AI models predicted World Cup match outcomes, Mistral emerged as the top performer among major AI models. According to Business Insider, startup ObSide conducted a real-world betting simulation with ChatGPT, Gemini, Claude, Grok, Mistral, DeepSeek, and Kimi. Each model switched to agent mode one hour before matches, researched team strength and injuries, and wagered portions of $10,000 based on real-time odds. By Thursday's semifinal closing, Mistral led, followed by GPT 5.5 and DeepSeek V4, while Claude was last and the only model recording losses. The test aimed to evaluate judgment in uncertain scenarios, similar to a previous case where ChatGPT matched human performance in an economic prediction tournament. This experiment highlights the limitations of decision-making AI despite standardized data and odds.