OpenAI’s o3 Dominates Elon Musk’s Grok 4 in AI Chess Tournament with a Clean Sweep
In a highly anticipated exhibition AI chess tournament, OpenAI’s large language model o3 decisively defeated Elon Musk’s AI model Grok 4 with a sweeping 4-0 victory in the final match held on August 7, 2025. The contest, hosted on Google’s Kaggle Game Arena, featured eight general-purpose AI language models competing against one another, including systems from Google, Anthropic, Moonshot AI, and DeepSeek.
Grok 4, developed by xAI, a company founded by Elon Musk, had been the strongest contender throughout the tournament, storming into the final round after outperforming other top models such as Google’s Gemini 2.5 Pro and Anthropic’s Claude 4 Opus. However, in the final, Grok 4 committed multiple critical mistakes early on, which opened the path for OpenAI’s o3 to dominate the match decisively.
According to chess world number one Magnus Carlsen, who provided live commentary with British Grandmaster David Howell, Grok 4’s gameplay was riddled with errors that no strong chess player would typically make. Carlsen estimated Grok 4’s chess skill rating to be around 800, while OpenAI’s o3 held a stronger estimated rating near 1200. Despite Grok’s solid performance up until the final, its inability to maintain strategic accuracy under pressure led to its clean 4-0 loss. In contrast, o3 displayed remarkable move accuracy with an average rate exceeding 90%, underscoring its superior reasoning and planning capabilities.
The tournament also marked a milestone as the first significant competition between general-purpose large language models testing their strategic reasoning abilities in chess, a game long dominated by specialized AI such as IBM’s Deep Blue and Google’s DeepMind. Experts noted that the event offered valuable insights into how these broad AI models think and reason in complex environments.
Hikaru Nakamura, the world’s number two chess player, observed that Grok seemed ‘nervous and anxious’ in the final round, likely contributing to its uncharacteristic mistakes and piece losses—errors that rarely appeared in its earlier matches against Google’s Gemini models.
Elon Musk responded on social media by downplaying Grok 4’s chess prowess, stating that chess capability was merely a “side effect” of xAI’s efforts, which focused minimally on chess during the AI’s development. Meanwhile, OpenAI celebrated the victory as validation of its advances in AI reasoning and strategic thinking through the o3 model.
OpenAI’s success in this AI chess tournament comes on the same day it announced the launch of its 11th generation large language model, GPT-5, although the tournament final featured the previous generation o3 model. This win not only demonstrates the growing sophistication of AI in mastering complex games but also revives the historic rivalry between Elon Musk and Sam Altman, OpenAI’s CEO, both of whom co-founded OpenAI a decade ago before going separate ways with competing AI ventures.
The tournament ended with Google’s Gemini 2.5 Pro securing the bronze medal by defeating OpenAI’s o4-mini, further highlighting the increasing competition in AI development from major tech companies. As AI continues to evolve, experts expect future tournaments to test these models in even more complex strategic games, hinting at a new era where general-purpose AI may challenge professional human players in various domains beyond chess.