OpenAI o3 Beats Grok 4 4-0 to Capture Kaggle AI Chess Title

OpenAI’s general-purpose LLM o3 dominated Kaggle’s AI chess competition, sweeping xAI’s Grok 4 4-0 in the finals. This marks the first time an untrained LLM has shut out an opponent in a major AI chess showdown. Over three elimination rounds, o3 delivered consecutive 4-0 victories, including a semifinal win over its own o4 mini. Experts highlighted o3’s minimal errors and strong inference skills. Chess grandmaster Hikaru Nakamura praised o3’s accuracy, while former world champion Magnus Carlsen likened Grok 4’s play to an amateur, assigning it an Elo rating around 800 compared to o3’s 1200. The event underscores the performance gap between general-purpose LLMs and specialized engines like Stockfish, which maintains an Elo above 3600. The AI chess face-off reveals that LLMs excel in reasoning tasks but still trail in deep search and computational depth. This outcome may influence future AI research priorities and model development in gaming and other inference-driven applications.