US wins AI chess tournament

AI Kimi K2 from China was disqualified for repeatedly making illegal moves - Photo: chess.com

With an absolute 4-0 victory, Gemini 2.5 Pro, o4-mini, Grok 4 and o3 have excellently advanced to the semi-finals of the AI chess tournament, after defeating Claude 4 Opus, DeepSeek R1, Gemini 2.5 Flash and Kimi k2 respectively.

In the most notable quarter-final, Moonshot AI's Kimi K2 model (China) suffered a disastrous 0-4 defeat against o3, LLM of OpenAI - the developer of ChatGPT.

All four games ended quickly in less than eight moves, as Kimi K2 repeatedly made illegal moves.

For example, in the third game, despite correctly understanding the position when o3 checked back, Kimi K2 still could not find a legal move in all four attempts, and was forced to concede defeat. The percentage of moves that matched o3's Stockfish tool was up to 100%, showing that the difference in skill level was too great.

The other Chinese representative, DeepSeek, did not fare much better, losing 0-4 to OpenAI's o4-mini. Although they did better than their compatriots by holding out in the first game, DeepSeek still made mistakes and was checkmated in the following games.

However, the most impressive character in the quarterfinals was Grok 4, a model from billionaire Elon Musk's xAI Company. Grok 4 easily defeated Google's Gemini 2.5 Flash with a score of 4-0. With the ability to punish every mistake of the opponent, Grok 4's move accuracy rate reached the highest level of the round, approximately 97.5%.

Commenting on the match, world No. 2 player Hikaru Nakamura was surprised: “Grok 4 is definitely the strongest LLM in this tournament. The level gap between it and the other models is not small.”

This comment was reinforced when Mr. Musk quickly re-shared the image of Nakamura's comment on the social network X, along with the confident comment: "This is just a side effect. xAI spends almost no time on chess."

Chess player Nakamura said Grok 4 was completely "out of the game" at the AI tournament - Photo: screenshot

On Google's side, although Gemini 2.5 Flash was eliminated, their remaining representative, Gemini 2.5 Pro, had a convincing 4-0 victory over Claude 4 Opus of Anthropic Company, affirming its position in the tournament.

The semi-finals will take place at 0:30 on August 7 (Vietnam time). The first semi-final is a high-level confrontation between Grok 4 and Gemini 2.5 Pro. The remaining match is a dramatic "OpenAI derby" between o3 and o4-mini.

TUAN LONG

Source: https://tuoitre.vn/my-thang-tuyet-doi-tai-giai-co-vua-danh-cho-ai-20250806111234074.htm