Anthropic's recent Claude Opus 4.7 and 4.6 releases, which incorporate extended reasoning modes, have driven scores on Humanity’s Last Exam into the mid-to-high 30s on the Scale AI leaderboard, narrowing the gap with leaders like Gemini 3.1 Pro Preview at around 44%. This progress stems from targeted improvements in handling graduate-level questions across math, science, and humanities, where earlier Claude versions scored below 20%. Competitive pressure from OpenAI’s GPT-5 series and Google’s Gemini updates continues to accelerate capability gains, though models still trail expert human performance near 90%. With the June 30 deadline weeks away, any surprise model iteration or benchmark-optimized training run could shift outcomes, while current trajectories suggest incremental rather than breakthrough gains remain most likely.