Artificial Analysis added new coding and scientific-agent tasks to Intelligence Index v5 and based 60% of the index on private tests or held-out answers. At the overhaul’s announcement, the gap between the leader and Kimi K3.1 Pro widened from 3 to 12 points, although that model’s weights had not yet been released. Knowledge work, coding, science and general capabilities each receive a 25% weight. Score changes therefore reflect the test mix as well as model capabilities; numbers from different versions and rankings alone are insufficient to judge cost-effective suitability.