OpenAI’s October 2026 releases of GPT-6 Astra, Sol, and Luna have driven recent gains on Humanity’s Last Exam (HLE), a 2,500-question expert benchmark testing frontier reasoning across disciplines. These models now post mid-50s scores on standard leaderboards such as Artificial Analysis and Scale AI variants, with subsets showing 82.9% when tools are permitted, lifting trader consensus for a 55%+ outcome to roughly 86% implied probability. Anthropic’s Claude Opus 5.5 and Fable 5.1 lead many full HLE tables at 61–65%, while Moonshot’s Kimi K3.1 Pro reaches 70% with tools, underscoring the narrow competitive gap. Further OpenAI updates or refined evaluations before year-end remain the main swing factors for 60%+ thresholds.
Experimental AI-generated summary referencing Polymarket data. This is not trading advice and plays no role in how this market resolves. · UpdatedView resolved

Beware of external links.
Beware of external links.
Frequently Asked Questions