Anthropic’s rapid release cadence for its Claude Opus series, including the September 22 launch of Opus 5.5, has positioned the family as a consistent leader on Humanity’s Last Exam variants, with the newest model topping HLE-with-tools leaderboards at 67.7% and related AA-HLE scores at 61.4%. The benchmark, a 2,500-question expert-authored set spanning math, physics, and multimodal reasoning that remains far from saturation, saw a refined HLE-Diamond subset released the same day, where GPT-6 Astra edged Opus 5.5 on the no-tools version. This competitive dynamic, alongside OpenAI and other labs pushing frontier models, makes HLE a focal point for demonstrating genuine capability gains rather than benchmark saturation. Traders should watch for Anthropic’s next Opus iteration timing, any official HLE-specific announcements, and whether internal safety or capability thresholds delay or accelerate its debut relative to rivals.