AI Model Benchmarks: Where Top Performers Excel and Struggle

AI Model Benchmarks: Where Top Performers Excel and StruggleWe analyzed performance across 1,276 benchmark tests spanning multiple AI organizations and task types. The data reveals stark differences: coding tasks prove hardest (42.6% average), while commonsense reasoning is easiest (92.4%), and xAI, DeepSeek, and Anthropic currently lead the rankings. Most models cluster around strong performance, but a handful lag significantly behind.We analyzed performance across 1,276 benchmark tests spanning multiple AI organizations and task types. The data reveals stark differences: coding tasks prove hardest (42.6% average), while commonsense reasoning is easiest (92.4%), and xAI, DeepSeek, and Anthropic currently lead the rankings. Most models cluster around strong performance, but a handful lag significantly behind.View this data story on Datastam