OpenAI’s GPT-6 Astra currently trails Anthropic’s Claude Opus 5.5 and Fable 5.1 on main Humanity’s Last Exam leaderboards, posting roughly 54–55% on Artificial Analysis HLE and similar verified runs as of early October 2026, while Claude variants reach 61–65%. A September HLE-Diamond subset result showed GPT-6 Astra at 82.9% with tools, highlighting how tool use and refined question sets can shift measured performance. Rapid frontier progress—roughly 30 points gained across labs in the past year—keeps the 2026 ceiling fluid, with additional GPT-6 variants or reasoning upgrades expected before year-end. Traders watch official agi.safe.ai resolutions and any new model cards for threshold moves on 55%+, 60%+, or higher bands.