OpenAI's o3 Model Dominates AI Chess Tournament, Defeats xAI's Grok 4 4-0

In a recent AI chess tournament, OpenAI's o3 model secured a dominant victory, defeating xAI's Grok 4 with a 4-0 score in the final match. The tournament, held on a prominent online platform, featured eight AI models competing in a three-day elimination format. Notably, o3 maintained a perfect 4-0 record throughout the tournament, including a dominant performance against its sibling model, o4 mini, in the semifinals.

Grok 4, on the other hand, showcased its strength by eliminating two of Google's top models, Gemini 2.5 Flash and Gemini 2.5 Pro, in the earlier rounds. However, its momentum was halted in the final match against o3. The tournament's unique challenge required AI models to play chess without any specialized training, relying solely on pre-match knowledge gathered from the internet.

The commentary team, which included international chess grandmasters, provided insightful analysis. One grandmaster noted that Grok 4 made several errors during the final match, while o3 played flawlessly. Another grandmaster, the world's top-ranked chess player, commented that the AI models' chess skills were comparable to those of a beginner, with an Elo rating of around 800. He further explained that these models excel at calculating material advantages but struggle with strategic play, likening their performance to someone who can gather ingredients but cannot cook a meal.

The tournament highlighted the evolving capabilities of AI in chess, contrasting with previous specialized programs like AlphaGo and Deep Blue, which were specifically designed for the game. Earlier this year, both Grok and ChatGPT were outmatched by Stockfish, a chess-specific AI system, in a tournament hosted by a chess grandmaster. This event underscored the limitations of general-purpose AI models in specialized tasks, despite their impressive overall capabilities.