OpenAI o3 Sets Record on ARC-AGI Test: Industry Debates Whether AGI Has Arrived

OpenAI o3 achieved a score of 87.5% on the ARC Prize benchmark, surpassing the human average of 85%, which some researchers consider a milestone for AGI. However, ARC-AGI founder François Chollet pointed out that o3 used massive computation, costing $1,000 per problem, fundamentally differing from human efficient reasoning. The debate over the definition of AGI has reignited academic discussions.