**Anthropic's Claude Opus 5.5 and Fable 5.1 currently lead most Humanity's Last Exam (HLE) leaderboards at 61–67% depending on the evaluation protocol, outpacing OpenAI's GPT-6 Astra and Google's Gemini 4 Argon.** HLE, a 2,500-question expert-authored benchmark spanning over 100 subjects and designed to resist saturation, has seen frontier large language models advance roughly 30 points in the past year through improved reasoning, scale, and tool use. Text-only runs on Artificial Analysis hover in the low 60s, while tool-augmented or HLE-Diamond subsets push higher, with GPT-6 Astra reaching 82.9% on one refined set. Trader focus centers on Q4 2026 model launches, reasoning enhancements, and whether any system crosses 70% on the official Scale AI or agi.safe.ai leaderboard before year-end, as competitive pressure from Anthropic, OpenAI, and open-weight contenders like DeepSeek continues to accelerate progress.