Sam Altman’s OpenAI Crushes Elon Musk’s Grok in AI Chess Championship

The Openi O3 Sam Altmana-which was outdated late last week by releasing GPT-5-December Elon Musk Grok 4 in four direct matches on Thursday won Google Kaggle Arena to play Ai chess exhibition.

You may think it was a super -complex spectacle of high -tech behemotes who put their explanation to a top test, but as an appetizer, let’s say that world champion Magnus Carlsen compared Both bots with a “talented child who does not know how the pieces move.”

The three-day tournament, which lasted from August 5 to August 7, forced Chatbot General Purpose-Da, the same ones that help you write an email and claim that you approach the human level-you play chess without any specialized training. There are no chess engines, without watching moves, just regardless of chess knowledge that they randomly absorbed from the Internet.

The results were about as elegant as you would expect from forcing a linguistic model to play a game on the board. Carlsen, who faced the finals, estimated that both AI games played at the level of occasional players who recently found out the 800 Elo rules. For context, he is probably the best chess player he ever lived, with Elo 2839 points. These AIs played as if they had learned the chess from the corrupt PDF.

“They oscillate between really, really good games and incomprehensible strings,” Carlsen said while broadcasting, after the game. At one point, after watching Grok walking his king directly into danger, he joked that he could think that they were playing the King of Brdo instead of chess.

Real games were like a master class in how not to play chess, even for those who don’t know the game. In the first match, Grok basically gave one of his important pieces for free, and then worsened things, tearing up more pieces while they were already behind.

The game two became even more strange. Grok tried to do what chess players call the “poisoned coastal” – a risky but legitimate strategy where you grab an enemy advocate that looks free but not. Apart from Grok completely grabbed the wrong pledgee, the one who apparently defended. His queen (the most powerful piece on the board) was immediately trapped and trapped.

Until the third game, Grok built what looked like a solid position – good control of the position, without obvious dangers and in the basis set to help you win the match. Then in the middle game he basically threw the ball directly to the opponent. He lost a piece after a piece in a fast succession.

That was actually weird, since before the game against O3, Grok was a pretty strong candidate, showing solid potential – so much so that he was praised by the great master chess by Hikar Nakamur. “Grok is easily the best so far, it’s just objective, easily the best.”

The fourth (and last) game provided the only true suspension. Openai’s O3 made a massive mistake at the beginning of the game, which is a great danger in any reasonable match. Nakamura, who was running the match, said that there were still “a few tricks” for O3.

He was right – O3 went back to win his queen and slowly squeezed out the victory, while the gigs play Endgame fell apart like a wet cardboard.

“Grok made so many mistakes in these games, but Openi is not,” Nakamura said during his Live Stream. This was a pretty turnaround from earlier a week.

Time could not be worse for Elona male. After the tomb of the strong early circles, he announced on X that his chess skills were just a “side effect” and that Xai was “almost no effort on chess.” This proved to be an underestimation.

Before this “official” chess tournament, international master Levy Rozman hosting your own tournament Earlier this year with less advanced models. He respected all the moves recommended by Chatboti, and the whole situation ended up as a complete mess with illegal moves, pieces of pieces and wrong budgets. Stockfish, Ai built especially for chess, eventually won the Chatgpt tournament. Altman’s AI was aligned with Muskovo in the semifinals, and Grok lost. So it’s 2-0 for herself.

However, this tournament was different. Each bot was given four chances to make a legal move – if they failed four times, they automatically lost. This was not hypothetical. In the early circles, Ais tried to teleport the pieces over the board, revive the dead pieces and move the vocabulary to the side as if they had played a version of the chess on the fever they had invented themselves.

They are disqualified.

Google’s twins grabbed third place by defeating another Openi model, saving some dignity for the tournament organizers. This bronze medal match contained a particularly absurdly drawn game in which both AI -A completely won positions at different points, but could not think of how to finish.

Carlsen pointed out that AI -and are better in counting captured pieces than they really delivered to checkmates – they understood the material advantage, but not how to win. It’s like great in collecting ingredients, but unable to cook a meal.

These are the same AI models that technicians claim to be approaching human intelligence, threatening white collar jobs and revolutionize how we work. Still, I can’t play a plaque game that has existed for 1500 years without trying to cheat or forget the rules.

So, it is probably safe to say that we are sure, and for now it will not take control of humanity.