OpenAI’s o3 Beats Grok 4 to Win AI Chess Tournament

Just a day before OpenAI’s o3 reasoning model was officially retired from ChatGPT on August 8 in favour of GPT-5 with built-in reasoning capabilities, it won Kaggle’s AI Chess Exhibition Tournament. The model beat xAI’s Grok-4 model to win the finals 4-0. 

The model reached the finals, beating the Kimi K2 Instruct model in the quarter-finals and the company’s own o4-mini model with the same scoreline. 

“Up until the semifinals, it seemed like nothing would be able to stop Grok 4 on its way to winning the event,” said Chess.com, the world’s leading platform for online chess. 

“But the illusion fell through on the last day of the tournament. The chatty o3 simply dismantled its mysterious opponent with four convincing wins. Grok's play was unrecognisable, blundering soon and often. And for the most part, o3 showed no mercy,” the platform added in its analysis of the game. 

In the match for the third place between Google’s Gemini 2.5 Pro and the o4-mini, the former model took the bronze medal with 2.5-1.5.

The tournament took place in Kagle’s Game Arena, a benchmarking platform where models from Google, Anthropic and OpenAI compete in game-based environments. “Games provide a clear, unambiguous signal of success: you either win, lose or draw. Their structured nature and measurable outcomes make them the perfect testbed for evaluating models,” Kaggle said in a blog post. 

Regarding chess, Kaggle said the decision by the model stems from internal logic rather than just memorisation. Models must adapt quickly, recover from mistakes and exploit opportunities, “like a human grandmaster”, added Kaggle.  

Full details of the game can be found here.