AI Predicting Premier League Results: Claude Opus 4.6 Tops, Grok Ranks Last

Recently, AI startup General Reasoning released a report testing 8 mainstream AI models on predicting 2023-24 English Premier League results and formulating betting strategies. Each model received an initial fund of 100,000 pounds for three simulation tests.

Anthropic's Claude Opus4.6 performed best with an average loss of 11.0% and final fund of 89,000 pounds; Grok ranked last, losing all funds once and failing to complete tasks twice with zero average fund; OpenAI's GPT-5.4 had an average loss of 13.6%, while Google's Gemini3.1 Pro showed large fluctuations—43.3% average loss but once achieved a 33.7% return.

The study noted AI overall lags behind humans, and industry static tests can't reflect real-world complexity. Additionally, Grok may get more enterprise applications as Musk asks banks involved in SpaceX's upcoming IPO to subscribe to Grok.