OpenAI has boasted that it defeated Elon Musk-backed Grok in the final of a high-stakes AI chess competition, winning the bragging rights in the latest twist of an escalating battle between the two companies. The competition was held on Google-owned Kaggle and included eight large language models developed by OpenAI, xAI, Google, Anthropic, and Chinese developers DeepSeek and Moonshot AI.
These models, in contrast to the classical chess engines like Stockfish or AlphaZero, were not designed to play chess but to cover an extensive variety of everyday challenges. However, the competition demonstrated their flexibility, planning, and skills to reason in situations of rules and complexity.
The O3 model of OpenAI won the final against Grok 4 of xAI. Google Gemini finished in third place, defeating another entry by OpenAI. Although Grok 4 had seemed strong in the initial rounds, it played terribly in the final, making unusual mistakes, such as losing its queen multiple times.
Chess.com writer Pedro Pinhata said, “Up to the semi-finals, it appeared that Grok 4 could not be defeated.” The illusion collapsed on the final day of the tournament. Hikaru Nakamura, a chess grandmaster who livestreamed the event, observed that Grok had a tendency to blunder, whereas o3 was very precise.
Musk dismissed the loss, writing on X that xAI had put “almost no effort on chess” and that the tournament run was therefore incidental to its core design. Nonetheless, the loss brings out the minor yet considerable variations in the form in which AI models learn and implement strategy when under pressure.
Chess has held a long history of measuring the capabilities of computational reasoning. From Deep Blue’s defeat of Garry Kasparov in 1997 to the defeat of Go champions in the late 2010s by Google DeepMind’s AlphaGo, these games have served as milestones in the history of AI. It is not yet the case that large language models can compete with dedicated engines, but competitions such as this one demonstrate their growing ability to do things other than text and conversation.
To OpenAI, the victory underlines the multifaceted nature of O3 and can act as a PR coup in its competition against Musk’s xAI. To the AI industry, it is an indication that even general-purpose models are improving in their ability to think multiple moves ahead, not only on the chessboard.