(Bloomberg) – In the dispute to see which artificial intelligence model is the best in chess, the final was between and two technology chiefs who have repeatedly faced each other over the future of AI.
The goal -organized elimination tournament took place this week, with teams from the US and China putting their great language models against each other. In addition to Musk and OpenAi de Altman ,, Google, Anthropic and Moonshot Ai also presented their models.
Also read:
Take your business to the next level with the help of the country’s leading entrepreneurs!
Despite the pompous name of the tournament -Ai Chess Tournament -, the level of chess is far from the levels achieved by game -dedicated programs, such as the Deepmind Alphazero, who has learned to play plaid, Shogi and GO alone at an unprecedented level.
Competitors, including Finalists Grok 4 and OpenAi O3, are known as LLMs (General Language Models), used to write, program and help college students to meet deadlines.
Still, given the animosity between Musk, 54, and Altman, 40, the winner will take home a great reason for pride. The two heads of technology founded OpenAi a decade ago, then Musk launched a competing AI startup and started a lawsuit to prevent OpenAI’s restructuring in an entity for profit.
Continues after advertising
In February, he gathered a wealthy allies for an unreshed and unsuccessful offer of $ 97 billion by the non -profit organization that controls the chatgPT. Altman, in response, criticized the offer, mocked musk AI products and called him a “bully.”
“This is a side effect, by the way. It has spent almost no effort on chess,” posted Musk on X. The same reasoning applies to all models that play in the tournament.
Google is promoting the event to mark the launch of Kagle Game Arena, where IA models will compete in multiple games – essentially an AI Olympics. The models will be classified based on factors such as adaptability and complex reasoning capacity, and will also explain their thinking while playing.
Continues after advertising
The conclusion so far is that the models are bad in chess, eventually losing control of the game and making mistakes that cost parts. If a model attempts four consecutive illegal movements, he loses the match. This happened a lot in the first round. The Kimi K2, developed by MoonShot Ai, based in China, could not pass from eight movements in any match, after failing to find a nice movement.
In the semifinals, Grok 4 defeated Google’s Gemini 2.5 Pro tiebreaker, while OpenAi O3 crushed OpenAi O4-Mini (a smaller and supposedly more agile version of O3) for four zero games.
© 2025 Bloomberg L.P.
