Artificial Analysis launches a leaderboard for HLE-Diamond, the refined Humanity's Last Exam

Artificial Analysis launched its HLE-Diamond leaderboard, its own evaluation of the refined 1,000-question version of Humanity's Last Exam, as part of v5 of its Intelligence Index. CAIS and Scale AI released HLE-Diamond after a year of cleaning and refining the questions.

These are Artificial Analysis's own measurements:

• HLE-Diamond labels every question as reasoning or knowledge. Models above 50% all score 16–35 points higher on reasoning questions than on knowledge ones. GPT-6.1 Sol (max) has the widest gap among the leaders, with 76% on reasoning and 41% on knowledge.
• Scores rise with higher effort settings, and so does cost. Claude Opus 5.5 gains 26 points from low effort (42%) to max (68%) at about 35x the cost per task; GPT-6 Astra gains 19 points from low (45%) to max (64%) at 11x the cost per task.

Scores, token use and cost per task for each model are on the leaderboard.