Recent releases of advanced Claude models, including Claude Fable 5 on June 9 and Opus 4.8 variants, have driven strong trader sentiment by pushing Anthropic's frontier large language models to leading or near-leading positions on Humanity's Last Exam (HLE), a 2,500-question benchmark testing graduate-level expertise across math, sciences, and humanities. Current top Claude variants post scores in the 53-65% range on public leaderboards, outpacing earlier 2025 results below 10% and competing closely with GPT-5 series and Gemini previews amid rapid iteration cycles. Competitive pressure from rival labs, combined with Anthropic's focus on reasoning depth and agentic capabilities, supports momentum, though HLE's held-out questions and calibration demands introduce uncertainty around last-minute gains before the June 30 cutoff.