An SEO specialist conducted an experiment to test the reasoning capabilities of six popular large language models on a chess puzzle. The puzzle required a mate in two moves and was presented via text coordinates. Only Grok was able to correctly solve the puzzle, identifying the correct first move and subsequent continuations. Other models like Gemini and ChatGPT came close, finding the correct first move but failing to calculate the second, while Claude and Microsoft Copilot made incorrect first moves, and DeepSeek generated illegal moves. AI
IMPACT Highlights current limitations in LLM's structured reasoning and strategic planning capabilities, even for advanced models.
RANK_REASON The item tests the capabilities of existing LLMs on a specific task, rather than announcing a new model or significant research.
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
