Artificial Analysis launched the HLE-Diamond leaderboard on October 27, 2026, featuring 1,000 questions combining reasoning, knowledge and 272 visual items. Google’s Gemini 4 Carbon ranked first with an overall accuracy of 69.1%, including scores of 79.6% in reasoning and 58.6% in knowledge. Anthropic’s Claude Opus 5.5 finished second at 67.9%, trailing the leader by only 1.2 points, while roughly 13% of the model variants tested achieved more than 50% accuracy. The best open-weight models remained more than 30 points behind the leading proprietary models, putting their scores below 40%. HLE-Diamond accounts for 10% of the Artificial Analysis Intelligence Index v5, and Gemini 4 Carbon was announced on October 28, 2026, with introductory pricing of $2 per million input tokens and $10 per million output tokens.
Source: Read the original article
