OpenAI, the creator of ChatGPT, has defeated Elon Musk’s Grok in the final of a tournament to crown the best chess-playing Artificial Intelligence (AI).
Historically, tech companies have often used chess to assess the progress and capabilities of a computer, with modern chess machines practically unbeatable even against the best human players.
But this contest did not involve computers specifically designed for chess – instead, it was held between AI programs designed for everyday use.
OpenAI’s o3 model emerged undefeated in the tournament, defeating xAI’s Grok 4 model in the final, adding fuel to the ongoing rivalry between the two companies.
Musk and Sam Altman, both co-founders of OpenAI, claim their latest models are the smartest in the world.
It is reported that Google’s Gemini model took third place in the tournament after defeating another OpenAI model, according to BBC.
However, while these AIs are talented in many everyday tasks, they are still improving in chess – with Grok making a number of mistakes during play, including repeatedly losing its queen.
“Up until the semifinals, it seemed as though nothing could stop Grok 4 on its path to victory,” said Pedro Pinhata, a writer for Chess.com.
“Despite some moments of weakness, X’s AI appeared to be by far the strongest chess player… But the illusion fell apart on the final day of the tournament.”
He said Grok’s “unfamiliar” and “flawed” play allowed o3 to claim a series of “convincing wins.”
“Grok made so many mistakes in these games, but OpenAI didn’t,” said chess grandmaster Hikaru Nakamura during his live stream of the final.
Why is AI playing chess?
The AI chess tournament was held on Kaggle, a Google-owned platform that allows data scientists to test their systems through competitions.
Eight large language models from Anthropic, Google, OpenAI, xAI, as well as Chinese developers DeepSeek and Moonshot AI, faced off against each other during the three-day Kaggle tournament.
AI developers use well-known tests as benchmarks to evaluate their models’ abilities in fields such as reasoning or coding.
As complex, rules-based strategy games, chess and Go have often been used to measure a model’s ability to learn how to best achieve a specific outcome – in this case, outperforming opponents to win.
