The Kaggle AI International Chess Tournament has concluded, with the untrained o3 defeating Grok 4 with a score of 4-0, demonstrating strong reasoning abilities. (Background: Musk threatened to sue Apple: App Store rankings show monopolistic behavior, maliciously suppressing Grok) (Context: Grok 4 is now available for free, Musk's xAI faces off against GPT-5) Recently, the results of the 'AI Chess Showcase' held by Google's Kaggle were announced on August 14, where OpenAI's general large language model o3 swept xAI's Grok 4 with a score of 4-0, winning the championship and becoming the first LLM to completely shut out an opponent without specialized training. The event featured 8 AI participants over three days, culminating in knockout rounds. Highlights of the language model competition According to OpenTools.ai, o3 achieved a series of 4-0 shutouts during its advancement, even eliminating its own lightweight version o4 mini in the semifinals. In contrast, Grok 4 often led in the early game but repeatedly 'lost the lead' (sacrificing the most powerful piece, the Queen) toward the end of the match. Chess grandmaster Hikaru Nakamura rated o3 as having 'very few mistakes' and noted that Grok 4 often encountered tactical self-destruction. Former world champion Magnus Carlsen described Grok's style of play as: 'like watching a child play chess.' He estimated Grok's Elo at about 800, and o3 at about 1200, both far below top human players or specialized chess AIs. Elo: A professional rating system (English: Elo rating system) created by Hungarian-American physicist Arpad Elo, is a method for measuring the level of various competitive activities, recognized as an authoritative standard for evaluating competitive levels today, widely used in chess, Go, football, basketball, and other sports. The highest Elo score in chess was set by Magnus Carlsen at 2882 points. General AI vs. Specialized AI The specialized systems like Stockfish rely on deep search and domain ratings, maintaining an Elo of about 3644 over the long term. General LLMs, however, learn from large-scale cross-domain corpora, with chess being just an extension of reasoning ability. Although o3 can defeat Grok 4, it was still outmatched by Stockfish earlier this year, indicating that general models still have gaps in stability and deep computation in chess games. Related reports Betting on OpenAI, Masayoshi Son has 'turned over a new leaf' Ethereum developers installed 'malicious AI plugins' and were hacked, clearing their crypto wallet in three days, with ten years of cybersecurity experience proving useless. a16z's latest insights: Is traditional e-commerce dead? AI-native platforms are redefining 'shopping.' "LLM Chess Tournament Concludes: OpenAI's o3 wins, xAI's Grok 4 shut out without winning a game" was first published in BlockTempo (the most influential blockchain news media).

#AI