The simmering rivalry between tech titans Sam Altman and Elon Musk has spilled onto the chessboard, with OpenAI’s latest model, o3, delivering a decisive and humiliating defeat to Musk’s Grok 4 in the final of a landmark AI chess tournament. The one-sided 4-0 victory for OpenAI marks a significant early win in a competition designed to test the strategic reasoning of general-purpose large language models (LLMs).
The tournament, hosted by Google’s Kaggle Game Arena, was the first of its kind, pitting some of the world’s most powerful AIs against each other in a chess showdown. The field included models from Google (Gemini), Anthropic (Claude), and others, but the final came down to the two models at the center of the AI gold rush: OpenAI’s o3 and xAI’s Grok 4. The match was a symbolic battle, given that Musk was a co-founder of OpenAI before leaving to launch his own rival company, xAI.
While Grok 4 had impressed in earlier rounds, with commentators praising its tactical instincts, the final was a masterclass in ruthless execution by o3. Grok’s performance was marred by a series of blunders that even casual chess players would recognize as disastrous, including questionable knight and bishop sacrifices and the inexplicable loss of its queen in multiple games.

Image source: pixahive.com
World number one chess player Magnus Carlsen, who was on hand to provide live commentary, was visibly stunned by some of Grok’s moves. At one point, he estimated Grok’s rating at a mere 800, a level that would place it on par with a beginner player. He later quipped that watching the final was “like watching kids’ games.” By contrast, Carlsen estimated OpenAI’s o3 to be a 1200-rated player, a solid intermediate level.
In the wake of the defeat, Musk took to his platform, X, to downplay the significance of the loss. “xAI spent almost no effort on chess,” he posted, calling Grok’s chess ability a mere “side effect” of its general reasoning capabilities. This statement, however, did little to blunt the public perception of a significant loss for his company.
While traditional chess engines like Stockfish and AlphaZero can play at a superhuman level, they are purpose-built for the game. This tournament was different, as it was designed to test the “emergent capabilities” of AIs not specifically trained for chess. The fact that the models could play at all was considered a feat in itself, and the final revealed a stark difference in their current strategic abilities.
For now, the victory gives OpenAI a significant boost in a high-stakes competition where public perception and performance are key. As these AIs continue to evolve, chess and other games are likely to become the new proving ground for the strategic and reasoning capabilities of these powerful new models.


















