OpenAI’s GPT-6 Astra and related variants currently post the highest publicly reported scores among its 2026 models on Humanity’s Last Exam, a 2,500-question expert-level benchmark testing frontier reasoning across science, math, and multimodal tasks. Recent September 2026 HLE-Diamond results showed Astra reaching 59.9% without tools and 82.9% with web and code tools, outpacing some Anthropic Claude runs on that subset while trailing Claude Opus 5.5 or Fable 5.1 on full-set or text-only leaderboards at 54–61%. Rapid capability gains of roughly 30 points year-over-year reflect scaling, post-training, and tool integration, though benchmark variants and verification methods create score dispersion. Traders watch for any late-year OpenAI releases or system-card updates before the December 31 resolution, as new model iterations or refined evaluations could shift the highest OpenAI mark.